Outdoor air pollution and lung cancer: Recent epidemiologic evidence

Outdoor air pollution has long been suspected of increasing the risk of lung cancer.1 Although many pollutants have been linked with adverse health impacts, the component with the greatest public health impact is probably PM, a complex mixture of airborne solid and liquid particles including soot, organic material, sulfates, nitrates, other salts, metals and biologic material. Many carcinogens, including a multitude of PAHs, adsorb to particles and can be deposited throughout the respiratory tract. In most Western countries, motor vehicle exhaust represents the most widespread source of air pollution, including PM.2 For research purposes, PM is usually subdivided into PM10 (inhalable particles), PM2.5 (fine particles) and PM0.1 (ultrafine particles). These PM size cuts generally represent different sources and display different physical and chemical properties. The physicochemical characteristics responsible for PM-associated toxicity are incompletely understood. For example, the relative contributions of particle size, number, mass, surface area and chemical composition have not been determined. For a given mass concentration (in μg/m3), the particle size cut will determine both particle number and surface area (e.g., for a mass concentration of 10 μg/m,3 a particle diameter of 2 μ corresponds to 1.2 particles/ml air with a surface area of 24 μ,2 while a particle diameter of 0.02 μ corresponds to 2.4 million particles/ml with a surface area of 3,000 μ2).3 A larger surface area potentially means greater exposure to carcinogenic agents, such as PAHs adsorbed to the particles. Long-term average concentrations of PM pollution are quite variable in Western countries, from <10 μg/m3 of PM10 (e.g., in northern Sweden) to >50 μg/m3 in southern Europe, with intermediate levels in the United States (for U.S. levels, see the website http://www.epa.gov/oar/aqtrnd97/brochure/pm10.html). Gaseous constituents of outdoor air pollution include nitrogen compounds (e.g., NO2), sulfur compounds (e.g., SO2), CO, CO2, ozone and other chemicals. Unlike gaseous pollutants, such as ozone or SO2, the composition of PM varies considerably over space and time. In this minireview, we focus on the health impacts of particulate and gaseous emissions from diesel and gasoline exhaust. The contribution of other sources of outdoor air pollution, in particular industrial emissions, is beyond the scope of the present review. ACS-II, American Cancer Society Study-II; AHSMOG, Adventist Health Study on SMOG; CI, confidence interval; PAH, polycyclic aromatic hydrocarbon; PM, particulate matter; RAL, relative adduct labeling; RR, relative risk; WBC, white blood cell. Exposure to air pollution has been related to total mortality in scores of studies.4, 5 Overall respiratory mortality has also been linked with elevated levels of air pollution.5 Prior to the publication of 3 American cohort studies, the evidence for a relationship between air pollution and lung cancer was somewhat equivocal, having been based on geographic comparisons, case-control studies and occupational studies of workers exposed to PAHs or diesel exhaust. Estimation of the effects of air pollution on lung cancer incidence is difficult for several reasons: (i) clinically detectable lung cancer takes many years to develop, (ii) long-term records of air pollution are required, (iii) information on other risk factors (potential confounders) is needed and (iv) measurements of both air pollution and confounding factors are often difficult and subject to errors of classification and change over time. Among the “pre-cohort” studies, urban–rural comparisons have shown that an “urban factor” is associated with a 10–40% increase in lung cancer mortality. An obvious limitation of these cross-sectional studies, however, is the impossibility of disentangling air pollution from other factors associated with urban life.6 Some follow-up investigations conducted in urban populations have provided stronger evidence of an urban–rural difference when smoking was taken into account, with rate ratios in the range of 1.1–1.5.7, 8 Case-control studies incorporating data on smoking and occupational exposures have shown an association between living in more polluted areas and the occurrence of lung cancer.9 In a large case-control study in Trieste, Italy, residence in the polluted city center was associated with increased risks of small cell (RR = 2.0, 95% CI 1.2–3.4) and large cell (RR = 2.6, 95% CI 1.2–5.3) lung cancer relative to residence in a suburban area.10 These estimates included adjustment for smoking and for occupational exposures to carcinogens. No increase in risk was seen for either squamous cell carcinoma or adenocarcinoma. The authors indicated that during the preceding 20 years most of the air pollution in the city center was attributable to transportation. A Swedish case-control study, involving 1,196 male lung cancer cases and 2,364 controls, detected associations between traffic exposure and lung cancer.11 Exposures from traffic (using annual average NO2 at each subject's residence as an indicator pollutant) and residential heating (using annual average concentrations of SO2) were estimated from 1950 through 1989. An estimated RR of 1.44 (95% CI 1.05–1.99) was found in association with more than 29.3 μg/m3 of NO2 (representing the highest decile of estimated average exposures over the period 21–30 years prior to the end of follow-up) compared to exposures lower than 12.8 μg/m.3 There was no association with estimated SO2 levels.11 During the last 10 years, several well-conducted cohort studies have been published. The main advantage of the cohort studies is that exposure is ascertained independently of and prior to the onset of clinically apparent disease, reducing the likelihood of bias. Three U.S. cohort studies were published in the 1990s.12, 13, 14 Long-term exposure was estimated in 2 studies from metropolitan annual average ambient concentrations; air pollution exposure contrasts were based on intercity concentration differences, while no information was available at the individual level.13, 14 A third study12 was conducted in California, which historically has had one of the most extensive air pollution–monitoring networks in the world. In this investigation, monthly air pollution concentrations were refined by interpolating data from the monitoring sites to the zip codes corresponding to the individuals' residential and work addresses. In all 3 studies, individual-level information was available on potential confounders or effect modifiers, which were generally assessed at the start of follow-up. The AHSMOG12 was based on 6,338 California Seventh Day Adventists followed from 1977 through 1992. The investigators reported substantial increases in RRs of lung cancer mortality among men in relation to long-term ambient concentrations of PM10 (RR = 3.36, 95% CI 1.57–7.19 associated with an interquartile range of 24 μg/m3) and SO2 (RR = 1.99, 95% CI 1.24–3.20 associated with an interquartile range of 3.7 ppb); however, as suggested by the wide CIs, these results were based on very few cases (n = 18 lung cancer deaths in men). While the mean ozone concentration was not associated with lung cancer incidence in men or women, there was an association of ozone with lung cancer risk in males when the exposure metric was formulated as the number of hours per year with elevated ozone concentrations (RR = 4.19, 95% CI 1.81–9.69 for at least 551 hr/year > 100 ppb ozone, controlling for pack-years of cigarette smoking, educational level and current alcohol use). Although the RRs for lung cancer in women (n = 12) for various PM10 metrics were also >1.0, all CIs included the null value. PM10 was measured only during the last 5 years of follow-up in this study; during most of the study, PM10 concentrations were estimated from measurements of total suspended particles. The Harvard Six Cities Study13 was based on 8,111 residents of 6 U.S. cities, followed from 1974 through 1989. Exposure was estimated on the basis of average levels of pollution over the risk period, assuming residential stability. RR estimates were adjusted for age, gender, smoking habits, body mass index and education. The total number of lung cancer deaths was reported as 8.4% of 1,429 (or 120). The difference in the long-term average PM concentrations between the most and least polluted cities was approximately 20 μg/m,3 corresponding to an RR of 1.37, i.e., an approximately 19% increase in risk per 10 μg/m.3 The third and largest U.S. investigation is the ACS-II,14 based on the mortality experience of approximately 500,000 adult men and women who were followed from 1982 through 1998. Participants were assigned to metropolitan areas of residence, and mean PM2.5 concentrations were compiled for each metropolitan area from several data sources. Personal information on risk factors (confounders or effect modifiers) was collected by questionnaire at enrollment. The study indicated a significantly increased mortality risk ratio for lung cancer (RR = 1.14, 95% CI 1.04–1.23) for a difference of 10 μg/m3 of PM2.5, controlling for age, gender, race, smoking, education, marital status, body mass, alcohol consumption, occupational exposure and diet. Therefore, 2 of the U.S. cohort studies, which are not subject to some of the principal limitations of previous geographic comparisons or case-control studies, suggest an excess risk of lung cancer of about 19% per 10 μg/m3 increment in long-term average PM2.5, after adjustment for the most likely confounding factors.13, 14 (Although the point estimates from the AHSMOG were also elevated, these are less precise, having been based on relatively few events.) Therefore, inhabitants of some southern European cities, e.g., could be exposed to air pollution–associated lung cancer risks about 75% higher than those among urban residents in northern Europe. However, this would be true only if one assumes a linear relationship between exposure and risk and that local effect modifiers (e.g., environmental tobacco smoke, diet or social class–related exposures) have little relevance (see below). In a reanalysis of the ACS data, linear vs. log-linear models were used to estimate air pollution–attributable lung cancer burdens at higher concentrations than those observed in the ACS-II.15 On the basis of the available evidence, one cannot rule out a linear exposure–response relationship at this time. The first published European cohort study examining long-term exposure to air pollution was conducted in the Netherlands.16 Conducting studies in Europe is extremely valuable for at least 3 reasons: (i) the 2 largest U.S. studies assessed only between-community spatial variations of air pollution and not within-community exposures, (ii) ambient levels of several key pollutants are more variable in Europe than in the United States and usually higher and (iii) European populations have a wide range of different exposures and living habits (particularly diet and smoking prevalence), which could act as effect modifiers. Enrollment in the Netherlands Cohort Study on Diet and Cancer started in 1986. At baseline there were 120,852 55- to 59-year-old adults living in 204 small towns and large cities throughout the Netherlands. A baseline questionnaire requesting information on active and passive smoking, occupation, education, nutrition and up to 4 residential addresses was administered. For the air pollution study, a subcohort of 5,000 subjects was used to estimate person-time within a case-cohort design. Mortality between 1986 and 1994 was studied in the subcohort. Exposure assessment was based on estimation of long-term exposure at the baseline home address and included assessment of NO2 and black smoke. The strongest predictor of lung cancer risk was identified as residential distance from a heavily trafficked street (within 100 m of a freeway or 50 m of a major urban street). While the greatest effect estimate for cardiopulmonary mortality was living near a busy street, the report by Hoek et al.16 indicates that there were too few cases of lung cancer to estimate the risks associated with this exposure metric. Risk of lung cancer was slightly elevated, but the estimate was based on relatively few cases (n = 60; RR = 1.06, 95% CI 0.43–2.63 for a 10 μg/m3 increment in exposure to black smoke; RR = 1.25, 95% CI 0.42–3.72 for a 30 μg/m3 increment in exposure to NO2). A second European study comes from Norway. Nafstad et al.17 studied lung cancer incidence among 16,209 40- to 49-year-old men living in Oslo, who were recruited in 1972–1973. Exposure assessment was based on measured concentrations of 2 gaseous air pollutants (NO2 and SO2) available from 1974 to 1995. Exposures in different areas of Oslo were contrasted. The population was followed through the Cancer Registry. Information on several potential confounders (smoking, social class, occupation, physical exercise) was available. The authors found a risk ratio of 1.08 (95% CI 1.02–1.15) for an increment of 10 μg/m3 of NO2, with an exposure–response relationship (Table I). This investigation has several strengths: (i) large size and adequate statistical power (>400 lung cancer cases); (ii) prospective examination of cancer incidence; (iii) absence of between-city heterogeneity (study was conducted in only one city); (iv) only 40- to 49-year-old men were studied; (v) exposure assessment could be refined with data on emigration and changes of address, which were not available in earlier studies; (vi) confounding was addressed by including individual information on age, smoking, social class, occupation and physical exercise. In a detailed reanalysis of both the Six Cities and the ACS-II cohort data, Krewski et al.18 found that level of education was inversely associated with RRs (Fig. 1). Degree of association with social class was important and could explain part of the observed geographic differences in risk estimates of the relationship between air pollution and lung cancer. Education attained modifies the effects of sulfate or particles. HS, high school. (From Krewski et al.18) It has been suggested that a synergistic effect exists between tobacco smoking and environmental pollution with respect to lung cancer induction.19 In a small case-control study of lung cancer among women in Athens,20 an effect of air pollution exposure was observed only among smokers, suggesting biologic effect modification. Samet and Cohen,21 reviewing the evidence for such a joint effect, indicate that the few studies that have examined this issue suggest the existence of a synergistic relationship between exposure to ambient air pollution and cigarette smoking. However, the magnitude of the joint effect is difficult to estimate because of exposure measurement error (for both cigarette smoking and air pollution) and the imprecision due to the small numbers of lung cancer cases occurring in nonsmokers. Despite numerous attempts, epidemiologic studies cannot disentangle the carcinogenic role played by single constituents of air pollution. The presence in ambient air of the myriad compounds recognized to be carcinogenic in animals supports the biologic plausibility of the nexus between air pollution and lung cancer. For instance, PAHs, which are carcinogenic in animals, are thought to play a role in occupational lung cancer and are emitted from a of sources of Although the role of these compounds in lung cancer in exposed to ambient air pollution and suggested that there is a of relative risks of lung cancer associated with exposure to from in cigarette to in workers to in residents of areas exposed to high levels of air pollution and in exposed to environmental tobacco smoke. These estimated risks the carcinogenic of the exposure but also to the of exposure to important component of air pollution is diesel a complex mixture of of and particles. The composition of diesel exhaust varies by and an emissions is The gaseous is of CO, of nitrogen and and Some of the gaseous are suspected or to cancer in particles at a rate about 20 greater than of the mass of diesel exhaust particles is μ in to the and of the lung The of small solid (or particles μ in many as different including a of recognized adsorb to diesel exhaust of the particulate The organic compounds such as and and Many of these and are and carcinogens. the evidence that both diesel and gasoline are carcinogenic to exposure to these also be carcinogenic to the respiratory of In an several studies of various occupational with potential exposure to diesel and gasoline including and The that the evidence for the of exposure to diesel exhaust was and that of exposure to gasoline exhaust was that diesel exhaust is probably carcinogenic to and in i.e., evidence in and in other gasoline exhaust was in i.e., evidence in and less than in other the cohort and case-control studies as as studies have addressed the issue of exposure to diesel exhaust in relation to lung cancer. the Health and the found that the epidemiologic data were in associations between exposure to diesel exhaust and lung cancer. from and 30 studies, have been in 2 These for the that occupational exposure to diesel exhaust is associated with increased risk of lung cancer. These studies, on found that long-term occupational exposure to diesel exhaust was associated with an approximately increase in the RR of lung cancer. The association was present when the various studies were by occupational study and smoking This association was when for the most important potential cigarette smoking, and when the issue of publication was studies also evidence of exposure–response on of likely exposures in the et also reported evidence of a diesel exposure–response to the epidemiologic of high concentrations of diesel exhaust concentrations > has in increased incidence of lung in studies of have been somewhat less while those in have been In with the results of diesel exposure of in which the particulate had been by no evidence of the gaseous is to a of have that of black particles results in a carcinogenic to that of diesel exhaust which to the of the particulate on both and evidence, the California identified diesel exhaust particulate as a air which in to such There were no measurements of diesel exposure of the study subjects in of the occupational epidemiologic several have been to the cancer risk for populations by diesel vehicle In has been estimated that at the mass concentrations of diesel particles in the μg/m3), between and lung per per year be In cancer risk assessment for air a represents an estimate of the carcinogenic of a chemical and to the number of excess cancer cases per million population exposed for a to a concentration of that in the United States have used the 95% CI on the likelihood estimate of the risk for this The California estimated the 95% CI of the cancer risks to be in the range of to 2.4 excess lung cancer cases per exposure to μg/m,3 that the more estimates are likely to be at the lower end of this to the data, the California risk estimate would between and 4 excess lung cancer cases per with the estimates by et Although there are large associated with these are with prior estimates of the impact of air pollution generally on lung While cancer risks in the not be that the greatest exposures to and In one of the diesel to the risk estimate for studies of was elevated, after on the individual studies adjusted for the smoking RR = 95% CI et estimated that in the could have an excess risk of lung cancer attributable to diesel exhaust exposure on the of which is about 2 of magnitude greater than the risks estimated for the population The epidemiologic and evidence diesel exhaust is more than that for in some in which there is exposure to both of there are of increased For example, as a 6 studies of and had an increased risk of (95% CI by the studies had adjusted for cigarette smoking little studies have to a of potential by which air pollution could or and studies have as an end in particular the of in which are related to exposure to aromatic compounds including the that such adduct also in such studies and biologic plausibility for a role of air pollution in the of lung cancer. studies in Europe have shown that levels of were higher among subjects heavily exposed to air This has been among exposed to urban residents in a area in the United and in In all these the more exposed subjects had differences from those less with of about 3 for the in the and in the A of workers exposed to traffic pollution and a of 100 residents were studied in were in at levels were significantly higher for traffic workers among = and current = In both urban residents to higher levels than those living in suburban areas and a with adduct levels highest in and in et observed that populations exposed to PAHs air pollution a relationship between levels of exposure and Among exposed the adduct level per of exposure was significantly lower than that measured among exposed at environmental exposure The exposure–response was observed in lung from exposed to be due to of or of at high levels of In exposed to PAH, a of the relationship between exposure and was On the basis of large cohort studies in the United States and Europe, there are for that air pollution increase lung cancer in with other risk such as active and passive smoking and occupational The of exposure to exhaust from both and a cancer risk is addressed in part by investigations of subjects exposed to these These studies indicate that diesel exhaust to the lung cancer data on occupational exposed to gasoline exhaust are Although several studies that in large cities, experience an increased risk of lung is difficult to disentangle the role of gasoline vs. diesel There are several to focus on the cancer risk associated with air pollution. the RRs associated with long-term exposures are small and difficult to population exposures are The impact on due to exposure to be quite the potential population cancer is when one the impacts of occupational exposures to motor vehicle and to population exposure to carcinogenic (e.g., diesel particles) will also be likely to other health impacts of exposure to air pollution. This report was by a of the European to investigation, for on an earlier of the