• lung cancer
• air pollution

Lung cancer accounts for 1.2 million deaths yearly worldwide, exceeding mortality from any other cancer in the developed countries.¹ The vast majority are caused by tobacco smoking, but environmental causes of cancer, including air pollution, have long been a concern also.² Outdoor air pollution has received particular attention lately as research has proliferated linking exposure, even at low ambient levels, to a wide range of adverse health effects including increased mortality and morbidity from non-malignant cardiovascular and respiratory disease and lung cancer. In response, international agencies such as the World Health Organization and governments in Europe, the US and Canada have reviewed existing air quality standards and, in many cases, moved to strengthen them. In the developed countries, where air quality has generally improved in recent decades, the scientific basis and public health efficacy of these actions have been questioned by industries whose emissions are regulated and others. In this context, reports linking air pollution and lung cancer are likely to attract attention and generate controversy. The publication of the paper by Nafstad and colleagues in this issue of Thorax is an occasion to consider both the contribution of this study to the evidence linking air pollution and lung cancer and what additional research may be needed.³

Exposure to outdoor air pollution has been associated with small relative increases in lung cancer in studies conducted over the past four decades.⁴ The epidemic of lung cancer emerging in the 1950s in the US and Europe motivated early research on the role of air pollution, including studies of migrants and urban-rural comparisons but, as the role of cigarette smoking became increasingly clear, interest in air pollution waned. However, recent prospective cohort and case-control studies which have taken into account tobacco smoking, as well as occupational and other risk factors, have continued to report increases in lung cancer associated with air pollution.^5–7 The American Cancer Society (ACS) study, which included 10 749 lung cancer deaths, reported that each 10 μg/m³ increment of fine particles (PM_2.5) was associated with an 8–14% increase in lung cancer.⁷ A causal interpretation is buttressed by other evidence. Urban air contains known and suspected human carcinogens such as benzo[α]pyrene, benzene, and 1,3-butadiene, together with carbon based particles onto which carcinogens may be adsorbed, oxidants such as ozone and nitrogen dioxide, and oxides of sulphur and nitrogen in particle form. Increased lung cancer has also been reported among workers occupationally exposed to components of urban air pollution such as polycyclic aromatic hydrocarbons and diesel exhaust.^8,9

In light of this evidence, the question is arguably not “Does air pollution cause some lung cancers?”, but rather “How many excess cases is it likely to cause?”. The answer to this question, and another—“Which pollutants, emitted by which sources, may be responsible?”—can potentially inform regulatory action to improve air quality and public health.

The current evidence suggests that lung cancer attributable to air pollution may occur among both smokers and non-smokers, and therefore both residual confounding and effect modification of the air pollution relative risk due to cigarette smoking must be considered. Nafstad et al³ report the relative risks of air pollution adjusted for cigarette smoking, but adjustment may not have controlled completely for potential confounding. The authors acknowledge that their study, like most other cohort studies, has information on cigarette smoking only at the beginning of the follow up period. The possibility that changes in tobacco use are correlated with exposure is difficult to rule out, although the association of lung cancer with air pollution was largely unaffected in the Six Cities study⁵ when longitudinal information on cigarette smoking was used in a recent reanalysis,¹⁰ and several case-control studies have found an increased risk following adjustment using time varying information.⁶ Several studies, including the one reported here by Nafstad et al, show an increased risk of lung cancer among self-reported never smokers, but the numbers in any single study are very small and the estimates imprecise. This also complicates efforts to estimate the numbers of cases in which both air pollution and smoking play a role. A study that includes large numbers of well documented never smokers may be the only approach that could address these concerns, if feasible.

Past approaches to exposure measurement also contribute to uncertainty in risk estimates. The ACS and Six Cities studies estimated the exposure of each participant based solely on long term average concentrations in their metropolitan area of residence. This approach may accurately reflect exposure to pollutants distributed homogenously over large areas for several decades but, if exposure at finer spatial and temporal scales is important, the estimates of relative risk may be inaccurate. Newer European and North American studies have begun to use spatial statistical methods to estimate individual long term exposure histories, linking residential histories, measurements of traffic density on nearby streets, and long term records of specific air pollutants, and can estimate how the size of the relative risk varies in time and space.^6,11,12 Hoek et al¹¹ observed larger relative effects on mortality from cardiopulmonary diseases as a result of air pollution near to major roads than from larger scale urban and regional air pollution, and Nyberg et al⁶ estimated the highest relative risks of lung cancer for exposure 20 years or more before diagnosis. By providing exposure estimates at the individual level, these studies also reduce the possibility of aggregate level (ecological) bias.^10,13

The effect of air pollution on lung cancer, fully manifest only decades after exposure, is a moving target. The emergence of cars and trucks as dominant modes of transportation and the decline in heavy industrial manufacturing in some developed countries since the mid 20th century, combined with effective air quality regulations, have changed both the nature of urban air pollution and patterns of human exposure. Over the time course of many recent lung cancer studies, decreases in larger respirable and fine particles as well as some gaseous pollutants and carcinogens have been documented,⁷ although concentrations of the fine, and arguably more toxic, particles may have declined to a lesser extent than other pollutants,⁵ increased in some locations,¹⁴ or changed their spatial distributions. Epidemiologists must rely on whatever components of the air pollution mix have been measured over extended periods, and consequently have reported associations of lung cancer with long term exposure to particles, ozone, sulphur dioxide, and nitrogen dioxide, but not known carcinogens. No mechanisms by which these pollutants per se cause cancer have been identified, and although some cancer biomarkers have been associated with air pollution exposure in non-smokers, they have not been used in large studies designed to estimate lung cancer risk. Nafstad et al³ and Nyberg et al⁶ used ambient concentrations of nitrogen dioxide and sulphur dioxide as surrogates for air pollution from mobile sources and residential heating, respectively. Each observed an increased risk of lung cancer associated with the nitrogen dioxide based indicator but not with the sulphur dioxide based indicator, but neither of these pollutants is specific to either source. As technological improvements and regulatory efforts continue to change the nature of air pollution, estimating current and future impacts on lung cancer will remain a challenge.

Exposure to air pollution is estimated to contribute to 62 000 lung cancer deaths per year worldwide—a large number of deaths, to be sure, but considerably less than the 712 000 deaths from non-malignant cardiac and respiratory disease attributable to air pollution.¹⁵ These impacts are largely borne by the populations of highly polluted cities in developing countries—roughly 60% of the world’s burden of air pollution attributed disease. In Chinese cities, where air pollution levels are many times greater than those in the cities of the developed West, outdoor air pollution may contribute to as much as 10% of lung cancer overall, and perhaps a larger proportion in non-smoking women. Unfortunately, because there is a lack of suitable studies in developing countries, these estimates are based on extrapolating the relative risk estimates from the ACS study to China, India, and other settings where differences in health status and the air pollution mixture introduce large uncertainties.

Opportunities to strengthen the scientific evidence on air pollution and lung cancer should be pursued, including in developing countries where the estimated health impact of air pollution and the need for accurate risk estimates are greatest. Studies should be designed to address, in addition to lung cancer, other arguably more important knowledge gaps such as the effect of long term exposure on the incidence of chronic non-malignant cardiorespiratory disease. Beginning large studies de novo would entail major financial and opportunity costs, so identifying existing cohorts, especially those with large numbers of non-smokers and for whom biological samples have been stored, may be the best option. Studies of outdoor air pollution and lung cancer in developing countries will need to account for past or concurrent exposures to indoor air pollution, particularly from use of coal for cooking and heating, a major cause of lung cancer in poor rural women in China and elsewhere,¹⁶ and changing patterns of tobacco smoking.