This study showed associations between different markers of air pollution from traffic at the residence and risk for lung cancer in a prospective cohort study. The results indicated stronger associations among nonsmokers, among participant with longer school attendance, and among those with relatively low dietary intake of fruit.
We used a large prospective cohort where information on potential confounding factors was collected at enrollment, that is, with no potential for recall bias. We ensured complete follow-up for cancer and vital status by using the population-based Danish Cancer Registry, the Danish Pathology Data Bank, and the Danish Central Population Registry. The Central Population Registry also provided information on residential addresses back to 1971, which we used in the exposure assessment.
The exposure assessment is a major challenge in studies of health effects of long-term exposure to air pollution. In the present study we used three markers for air pollution from traffic at the residences, which were moderately correlated (r between 0.43 and 0.53). We calculated the outdoor NOx level for all addresses over decades using a validated model requiring comprehensive input data, and the two other markers are simple, intuitively understandable measures of traffic at the residence at the time of enrollment, with one referring to presence of a major road within 50 m of the residence and the other summarizing the total traffic load within 200 m of the residence. Although the calculated NOx concentration is the marker that takes into account most factors influencing the long-term NOx concentration, to our knowledge no studies have determined whether such comprehensive modeling of the concentration at the front door in fact reflects personal exposure of the individuals living at the address better than simpler indicators for traffic in the neighborhood. To properly illuminate which of the three exposure markers best reflects personal exposure of persons living at the address would require measurements of personal exposure as the gold standard for comparison with each of the exposure markers.
Exposure over a long period, perhaps over a whole life, is probably relevant for the development of lung cancer, and the present study benefited from information on residential histories from 1971 onward as the basis for the assessment of exposure to NOx
. One of the few previous studies with information on exposure decades back in time indicated that the effect of air pollution on the risk for lung cancer is stronger after inclusion of a lag, that is, after disregarding exposure during the period closest to the diagnosis (Nyberg et al. 2000
). In the present study, however, the NOx
concentration calculated for the whole exposure period correlated strongly with both the NOx
concentration calculated after inclusion of a 10-year lag (r
= 0.98) and that calculated as an annual average for the address at the time of enrollment (r
= 0.86). Thus, our material would seem not to be suitable for investigating possible effects of timing of exposure. The high correlation between NOx
averages for all addresses since 1971 and the annual mean of NOx
at the residence at time of enrollment indicates that the exposure markers based on the address at the time of enrollment in the present study might reflect exposure for a much longer time period.
The dispersion models we used to assess NOx
levels at the addresses of study participants have been successfully validated (Berkowicz et al. 2008
; Raaschou-Nielsen et al. 2000
) and applied both in Denmark (Raaschou-Nielsen et al. 2001
) and in the United States (Jensen et al. 2009b
). Also, we used indicators for traffic near the residence. Such markers of air pollution concentrations are inevitably associated with some degree of uncertainty. We cannot see how this uncertainty could depend on development of lung cancer, however, and such nondifferential misclassification would only in rare situations create artificial associations (Dosemeci et al. 1990
Our finding of associations between indicators of air pollution from traffic and risk for lung cancer is in accordance with findings of previous studies showing effects of NO2
or proximity to traffic (Beelen et al. 2008
; Filleul et al. 2005
; Nafstad et al. 2003
; Nyberg et al. 2000
; Vineis et al. 2006
). A previous cohort study, conducted in Norway, that used NOx
as an indicator of air pollution (Nafstad et al. 2003
) showed a risk ratio for lung cancer of 1.36 (95% CI, 1.01–1.83) in association with ≥ 30 μg/m3
at the residence compared with < 10 μg/m3
. That is similar to the rate ratio of 1.30 (95% CI, 1.05–1.61) in association with ≥ 30 μg/m3
at the residence compared with < 17 μg/m3
that we found in the present study. Further, a previous case–cohort study combining data from three Danish cohorts and including 241 of the 592 cases in the present study showed an IRR of 1.37 (95% CI, 1.06–1.76) per 100 μg/m3
(linear trend) (Raaschou-Nielsen et al. 2010
), which is higher than the IRR of 1.09 (95% CI, 0.79–1.51) we found in the present study, although the CIs widely overlapped.
Two previous studies have investigated the risk for lung cancer in association with residence near heavy-traffic roads. Identical risk estimates for proximity to traffic cannot be expected in different studies because of different definitions of this exposure variable and differences in study populations. Nevertheless, Vineis et al. (2006)
found an odds ratio of 1.46 (95% CI, 0.89–2.40) and Beelen et al. (2008)
found an IRR of 1.11 (95% CI, 0.91–1.34), which are both comparable with the IRR of 1.21 (95% CI, 0.95–1.55) that we found in the present study.
Although associations have been found with NOx
in the present and previous studies, these single pollutants should be considered indicators for vehicle engine exhaust, which is a complex mixture including many carcinogenic and mutagenic chemicals (IARC 1989
has been shown to correlate closely with PM, especially the ultrafine fraction emitted from diesel engines in Danish streets (Hertel et al. 2001
; Ketzel et al. 2003
). It is difficult to disentangle the effect of single air pollutants in epidemiologic designs because they are part of complex mixtures, but it seems likely that for cancer risk PM from traffic emissions is most important. This comprises mainly PM with ultrafine size, large surface area, and absorbed polycyclic aromatic hydrocarbons, transition metals, and other substances causing oxidative stress, inflammation, and direct and indirect genotoxicity (Borm et al. 2004
; Moller et al. 2010
). However, although NO2
is not known as genotoxic, it cannot be excluded that NO2
can act as a tumor promoter in relation to diesel exhaust particles, as shown in an animal carcinogenicity study (Ohyama et al. 1999
). In the present study, we focused on air pollution from traffic, which is the major source of NOx
air pollution in Danish cities. Two other Scandinavian studies showed effects of traffic-related NOx
but not of heating and industry-related SO2
on the risk for lung cancer (Nafstad et al. 2003
; Nyberg et al. 2000
The relative risk estimates in the present study were substantially attenuated by adjustment for active smoking, consistent with smoking being a strong risk factor for lung cancer and smoking being associated with residence at locations with high air pollution levels (). However, we found stronger relative risks for lung cancer in association with air pollution among nonsmokers than among smokers, indicating that residual confounding by smoking is not the explanation for the observed associations between traffic pollution and risk for lung cancer. The higher relative risk among nonsmokers is in accordance with the results of previous studies (Beelen et al. 2008
; Nyberg et al. 2000
; Pope et al. 2002
; Yorifuji et al. 2010
). The lung cancer incidence rate is much lower among nonsmokers than among smokers (), and a higher relative risk among nonsmokers therefore does not necessarily correspond to a higher absolute risk. Thus, the 91% higher relative risk for the upper air pollution exposure group among nonsmokers would correspond to 27 excess lung cancer cases per 100,000 person-years (0.91 × 30 per 100,000 person-years), whereas the 21% higher relative risk among smokers would correspond to 57 extra cases per 100,000 person-years (0.21 × 270 per 100,000 person-years). Nevertheless, the wide CIs around the risk estimates indicate that the different relative risk estimates for nonsmokers and present smokers should be interpreted with caution.
The results of this and a previous study (Beelen et al. 2008
) indicate that an association between air pollution and risk for lung cancer might be present mainly among individuals with a low fruit consumption. An expert panel established by the World Cancer Research Fund and the American Institute for Cancer Research concluded in 2007 that fruits probably protect against lung cancer (World Cancer Research Fund/American Institute for Cancer Research 2007
). Among the possible mechanisms are the scavengers of free radicals and reactive oxygen species, as well as possible up-regulation of protective enzymes by constituents present in fruits, protecting against oxidative damage to the DNA (Moller and Loft 2006
). The suggested modification by fruit intake of an effect of air pollution on the risk for lung cancer might similarly be explained by the presence of oxidants and prooxidants in air pollution (Borm et al. 2004
; Moller et al. 2010
), which might lead to a higher risk for lung cancer mainly in individuals with low intake of antioxidants in fruits. A high intake of fruits might also protect against lung cancer by reducing the formation of DNA adducts from polycyclic aromatic hydrocarbons (Palli et al. 2004
; Peluso et al. 2000
). Fruit consumption might also be a marker of other characteristics, which might have contributed to the results.
The present study showed a stronger association between markers of air pollution and risk for lung cancer among individuals with ≥ 8 years of school attendance. The opposite tendency was observed in a previous study from the United States (Pope et al. 2002
), although the pattern was less clear in a recent analysis with a longer follow-up period (Krewski et al. 2009
). A Dutch study showed no clear modification of the association by educational level (Beelen et al. 2008
). It is obvious that educational level per se does not influence the risk for lung cancer. Instead, a variety of lifestyle and behavioral factors, other exposures, and overall health that are associated with educational level might influence lung cancer risk. Associations between educational level and risk (or protective) factors for lung cancer might well differ by country and time period, possibly explaining the heterogeneity of these results. The apparent effect modification in the present study might also be due to chance.
There is a partial overlap in the case series of this and a previous lung cancer study (Raaschou-Nielsen et al. 2010
). The present study, however, differs from the previous in many aspects, among which are the longer follow-up (and hence most new cases), the classical cohort design with exposure assessment for all 52,970 included cohort members, inclusion of two traffic exposure variables, and inclusion of ETS, occupation, and fruit intake as potential confounders or effect modifiers. Moreover, the classical cohort design of the present study facilitated calculation of absolute lung cancer rates in strata defined by potential effect modifiers, which qualifies the discussion about different effects of air pollution between smokers and nonsmokers. The overall conclusions of this and the previous study are very similar.
In conclusion, this study showed associations between risk for lung cancer and different markers of air pollution from traffic near the residence, in line with the weight of the epidemiologic evidence to date.