Most studies on long-term exposure to air pollution and lung cancer risk have investigated the association with lung cancer mortality. Fine particles were predictors of lung cancer mortality in the extended follow-up of the Six Cities cohort study1 and the American Cancer Society cohort study.2 Evidence that air pollution is a (moderate) risk factor for lung cancer mortality has also been found in several case-control studies,3,4 with effect estimates of the same order of magnitude as those estimated from cohort studies.5
Few studies have investigated the association between air pollution and lung cancer incidence.6–9 In the Adventist Health Study on Smog cohort, incident lung cancer was found to be associated with long-term exposure to ambient concentrations of fine particles and sulfur dioxide.6 This study was conducted among adult Seventh Day Adventists, who are largely nonsmoking. A nested case-control study within the European Prospective Investigation into Cancer and Nutrition, which analyzed nonsmokers, found associations between the nitrogen dioxide concentration and lung cancer incidence.7 Two studies in Scandinavia found elevated relative risks for traffic-related air pollution assessed by nitrogen dioxide concentration at the home address and lung cancer incidence.8,9
Recently, we reported on the association between traffic-related air pollution and cause-specific mortality, including lung cancer, in an ongoing Dutch cohort study.10 In the present paper, we report the effects of long-term exposure to traffic-related air pollution on lung cancer incidence in the same cohort. Cancer incidence data are preferred to mortality data in etiologic studies. Cancer registration in the Netherlands is both very complete11 and of high quality.12 Cases are coded by specially trained tumor registrars using national guidelines. Coders have access to pathology reports, clinical, and outpatient files. In case of doubt, the treating clinician is consulted.13 The cause-of-death registry contains information on all deceased Dutch citizens, as reported by the physician who established the death. This physician is not always informed on the complete disease history of the patient. The chances for misclassification (wrong primary site, missing primary site, or site of the metastasis as primary site) are much larger in the causes-of-death registry than in the cancer registry. The use of cancer registry data also allowed us to confine our analysis to the pathology-confirmed cases, reducing the possibilities of misclassification further. Further, due to a longer follow-up time for lung cancer incidence (11.3 years) compared with mortality (10 years), the lung cancer incidence analyses had more cases and more power.
The cohort has been described in detail.14 Briefly, the Netherlands Cohort Study on Diet and Cancer was initiated in September 1986 with the enrollment of 120,852 subjects (58,279 men and 62,573 women) aged 55 to 69 years living in 204 municipalities throughout the country. The study was designed as a case-cohort study, ie, cases are derived from the entire cohort, whereas the person-years at risk are estimated from a random subcohort (N ∼ 5000).15 This approach was chosen for efficient processing of the baseline questionnaire.
At baseline, all participants completed an 11-page questionnaire on dietary habits and other potential risk factors for cancer. For all participants, data from one machine-readable page of the questionnaire were entered at baseline (with information about smoking status). Further, for all participants information about sex and age at baseline was known. After recruitment, the entire cohort was followed-up for cancer. After the case-cohort design, the remaining 10 questionnaire pages (not machine-readable) were manually entered only for the emerging cases and the randomly selected subcohort. The exact residential address at baseline was available for all study participants, as were the previous cities of residence (not exact previous addresses).
Incident lung cancer cases (International Classification of Diseases for Oncology code [ICD-O-3] = C34) were identified by computerized record linkage of the entire cohort to the Netherlands Cancer Registry and the nationwide network and registry of histopathology and cytopathology in the Netherlands.16 The completeness of cancer follow-up was over 96%,17 with a follow-up period of 11.3 years (from September 1986 to December 1997). We excluded prevalent cancer other than skin cancer cases at baseline, resulting in 114,378 subjects and 2183 lung cancer cases available for analyses.
The study was approved by institutional review boards from Maastricht University and the Netherlands Organization for Applied Scientific Research. All cohort members consented to participation by completing the mailed, self-administered questionnaire.
Air Pollution Exposure Assessment
Details of the exposure assessment have been described previously.18 In summary, long-term exposure to outdoor air pollution at the geographical coordinate of the 1986 home address was estimated for all participants as the sum of regional, urban, and local air pollution contributions. Ninety percent of participants had lived for 10 years or longer in their 1986 municipality.19 Regional background concentrations were estimated using inverse distance weighted interpolation of concentrations measured at regional background sites in the National Air Quality Monitoring Network. The urban component was estimated using land-use regression models with concentrations for all regional and urban background monitoring sites as dependent variables. Predictor variables were population density and land use variables (residential or industrial). The sum of the regional and urban contributions was defined as background concentration. Background concentrations were estimated for nitrogen dioxide (NO2), black smoke, and sulfur dioxide (SO2). Average concentrations were estimated for 1976–1985 and 1987–1996 (in 1986 the Monitoring Network was reorganized, resulting in only limited days with valid measurements in that year). Correlations between estimated concentrations for different years were high, even over a period of 20 years (correlation coefficients >0.8).18
Local traffic contributions were characterized by traffic variables using a Geographic Information System (GIS) and a digital road network with linked traffic intensity data from 1986. We used (1) traffic intensity on the nearest road, (2) sum of traffic intensity in a 100 m buffer around each residential address, and (3) an indicator variable for living near a major road defined as “living within 100 m of a motorway and/or within 50 m of a local road with traffic intensity >10,000 motor vehicles per 24 hours (motor vehicles/24 hours).” Although absolute traffic intensities increased during the follow-up period, traffic intensity data obtained for different years were highly correlated (correlation coefficients >0.9), even over a period of 10 years.18 Further, quantitative estimates for the local component were estimated for NO2, black smoke and PM2.5 using data from field monitoring campaigns and regression models with traffic variables as predictor variables.18 We estimated no local traffic contribution for SO2 because there is virtually no traffic contribution to this pollutant. These local-component concentrations were added to the background concentrations, resulting in an overall exposure estimate for each pollutant.
Air pollution effects were analyzed for overall concentrations of pollutants and for a combination of background concentrations of the pollutants and traffic variables to identify effects of living near busy roads separately.
Relative risks (RRs) and 95% confidence intervals (95% CIs) were calculated for concentration and traffic variable differences between the 5th and the 95th percentiles of the distributions. For NO2 this was rounded to 30 μg/m3, for black smoke 10 μg/m3, for SO2 20 μg/m3, and for PM2.5 10 μg/m3. For traffic intensity on the nearest road, we used 10,000 motor vehicles/24 hours and for the sum of traffic intensity in a buffer of 100 m we used 335,000 motor vehicles/24 hours.
We initially conducted analyses in the full cohort using Cox proportional hazards models. Person-years were calculated for all participants from baseline until the date of lung cancer diagnosis, death, or end of follow-up. We adjusted for sex, age at baseline, and smoking status coded as never-smoker, ex-smoker, and current smoker separately for cigarette, cigar and pipe smoking. We further adjusted for area-level indicators assessed using GIS data from the Central Bureau of Statistics: percentage of persons with a low and with a high income at the neighborhood scale and the “COROP area scale.” “COROP areas” have been defined in 1970 by the Coordination Commission for Regional Research Program as geographic regions consisting of a central point (eg, a city) and the surrounding economic and social region. The Netherlands is divided into 40 such areas. We chose 2 scales for area-level socioeconomic status because life expectancy varies considerably by area,20 and the neighborhood scale does not capture such regional variations adequately. Low income was defined by the Central Bureau of Statistics as below the 40th percentile and high income as above the 80th percentile of the Dutch income distribution (Table 1). The adjusted full cohort analyses included 1940 cases.
For more complete confounder control (at the cost of some power), we also conducted case-cohort analyses that included only subcohort members and cases. Cases were enumerated from the entire cohort, whereas person-years for the entire cohort were estimated using the random subcohort of 4755 participants. Data were again analyzed with Cox proportional hazards models, but to account for additional variance introduced by sampling from the cohort, standard errors were estimated using the robust Huber-White sandwich estimator.21
In the case-cohort analyses, we adjusted for the variables chosen a priori for studying the association between air pollution and lung cancer incidence: sex; age at baseline; active cigarette, cigar and pipe smoking coded as current/noncurrent and number of cigarettes/cigars/pipes and number of years of smoking; passive smoking defined as whether the partner smoked; educational level in 3 categories: primary school, lower vocational education, and high school and higher; occupational exposure during the last occupation to biologic dust, mineral dust, and gases and fumes (coded as no, low and high exposure using the ALOHA-JEM coding scheme)22; alcohol consumption in 2 categories (0–30 and >30 g/d); dietary habits: intake of vegetables (continuous), fruit (in quintiles), and folate (continuous); and the area level indicators of socioeconomic status. The adjusted case-cohort analyses included 1295 cases.
As in other cohort studies of air pollution and lung cancer,6,23 we conducted subgroup analyses for sex, educational level (3 categories: primary school, lower vocational education, and high school and higher), and fruit consumption (tertiles: low [0–96.8 g/d], medium [96.8–191.8 g/d], and high [>191.8 g/d]). Further, we conducted subgroup analyses for cigarette smoking status (never/ex/current). The never-smoker group is of particular interest in not being subject to possible residual confounding by the number of cigarettes and number of years of smoking cigarettes. Subgroup analyses by sex and cigarette smoking status were assessed in the full cohort, and subgroup analyses for educational level and fruit consumption were assessed in the case-cohort (information about education and fruit consumption was available only in the case-cohort sample). Heterogeneity in relative risks across subgroups was tested using Cochran's Q test.24
We conducted sensitivity analyses in the case-cohort for black smoke and the traffic variables to evaluate the potential confounding effect of the number of cigarettes, cigars and pipe, and the number of years of smoking, because no information about this was available in the full cohort. Further, we conducted sensitivity analyses adjusting for occupational exposure to biologic dust, mineral dust, and gases and fumes during the longest-held occupation rather than the last occupation, as occupational exposure earlier in life may be more important than recent exposure. We also conducted case-cohort analyses adjusted only for the limited confounder model used in the full cohort analyses, and compared the RRs with those obtained with the complete set of potential confounders.
Data management was done using SPSS 12.0 (SPSS Inc, Chicago, IL) and statistical analyses were conducted using STATA statistical software 8 (StataCorp, College Station). GIS calculations were conducted using ArcInfo (ESRI, Redlands, CA).
During 11.3 years of follow-up, 2183 lung cancer cases were reported. For 98% of the subjects, a geographical coordinate at baseline was identified (n = 111,816 in full cohort). Table 1 shows the characteristics of the subjects in the full cohort for whom a coordinate was available. Lung cancer cases were older, predominantly men, and more likely to be current smokers. Lung cancer cases did not differ from non-cancer cases in the area-level socioeconomic status.
Air Pollution Exposure Data
There was a considerable contrast in air pollution exposure among the full cohort (Table 2). Distributions for the traffic variables were skewed. In the full cohort 5481 subjects (4.9%) lived within 50 m of a road with more than 10,000 motor vehicles/24 hours or within 100 m of a motorway. Average exposure for all exposure variables was higher for current smokers, but differences were small. The black smoke average concentration was 16.8, 16.6, and 16.3 μg/m3 for current, ex-, and never-smokers, respectively.
Distributions were similar for different time periods. For each pollutant the estimated concentrations between the periods 1976–1985 and 1987–1996 were highly correlated (correlation coefficients >0.9). The correlations between different pollutants within the same period were all >0.8, except for SO2 for which correlations were >0.6.18 Correlations of background black smoke with traffic intensity on the nearest road and sum of traffic intensity in a 100 m buffer were modest: 0.12 and 0.28, respectively. More detail on the exposure variables has been reported previously.18
Association Between Air Pollution Concentrations and Lung Cancer Incidence
Table 3 shows the relative risks for exposure to average black smoke, PM2.5, NO2, and SO2 concentrations during the period 1987–1996. In both the full cohort and case-cohort analyses, effect estimates for all pollutants were below one. Average concentrations for the period 1976–1985 were available only for NO2 and SO2. The RRs for NO2 were 0.91 (95% CI = 0.76–1.08) and 0.98 (0.69–1.40) for full cohort and case-cohort analyses, respectively. Corresponding figures for SO2 during the period 1976–1985 were 0.94 (0.84–1.07) and 0.96 (0.75–1.21). There were no differences between the RRs for the 2 periods, consistent with the high correlation of the concentrations for the 2 time periods.
The RRs for the traffic variables were slightly elevated for both the full cohort and case-cohort analyses (Table 3). Risk estimates of the traffic variables did not depend on the specific background exposure with which it was modeled simultaneously, so we present results from models with just 1 pollutant—black smoke concentrations for the period 1987–1996.
In the full cohort analyses, RR estimates for overall concentrations of black smoke and traffic variables tended to differ for never-, ex-, and current cigarette smokers (Table 4). There were no differences between the effect estimates for men and women for both black smoke concentrations and traffic variables (data not shown).
In the case-cohort analyses, no consistent differences in effect estimates were found for different educational levels (Fig. 1). Effect estimates tended to be higher for subjects with low fruit consumption, except for those living near a major road for whom no differences were found (Fig. 2).
Adjusted case-cohort analyses showed that there were no differences in the effect estimates for overall concentrations of black smoke for models that included (RR = 1.03; CI = 0.78–1.34) or did not include (1.01; 0.78–1.32) the number of cigarettes, cigars and pipes, and the number of years of smoking. There were also no differences between the effect estimates for the traffic variables.
There was no difference in the RRs for black smoke concentrations and traffic variables when including occupational exposure to biologic dust, mineral dust, and gases and fumes during the longest-held occupation in the model (for black smoke, RR = 1.03 [CI = 0.78–1.34]) compared with a model that included occupational exposure during the last occupation (for black smoke; 1.01 [0.76–1.33]).
The RRs of case-cohort analyses adjusted for the limited confounder model used in the full cohort analyses were 0.92 (CI 0.74–1.13) for black smoke, 0.98 (0.86–1.10) for the traffic intensity on the nearest road, 1.07 (0.80–1.43) for living near a major road, and 1.10 (0.83–1.22) for the traffic intensity in a 100 m buffer.
In the full cohort, we found no association of air pollution concentrations or traffic intensity with lung cancer incidence. Among people who had never smoked cigarettes, there were positive associations of black smoke concentrations and traffic variables with lung cancer incidence. No associations were found in the ex- and current cigarette smokers. Associations of black smoke and traffic variables with lung cancer incidence were suggestively stronger in subjects with low fruit consumption.
Stronger associations between air pollution and lung cancer in never-smokers have been observed in the American Cancer Society study, as well.2 Relative risks associated with a 10 μg/m3 increase in the concentration of fine particles were about 1.03, 1.09, and 1.13 in current, ex-, and never- smokers, respectively.2 In a cohort study from Oslo, RRs per 10 μg/m3 increase in the concentration of NOx were 1.20, 1.25, 1.06, and 1.06 for never-smokers, and current smokers smoking 1 to 9 cigarettes/d, 10 to 19 cigarettes/d and more than 20 cigarettes/d, respectively.9 In a case-control study from Stockholm, the RR for NO2 above 29.3 μg/m3 (the 90th percentile) was similar in never-smokers (RR 1.68), former smokers (RR 1.58) and those smoking 1 to 10 cigarettes per day (RR 1.36), but was smaller in current smokers smoking more than 20 cigarettes per day (RR 0.74).8
The lack of association found for ex- or current cigarette smokers in our study might be because the effect may be too small to measure in comparison with the (strong) association between cigarette smoking and lung cancer incidence. In the full cohort analyses, no information was available about the number of cigarettes, cigars and pipe, the number of years of smoking, the type of tobacco used, and the year when former smokers stopped smoking, which are all potentially important confounders. Sensitivity analyses in the case-cohort dataset showed small differences in effect estimates for air pollution variables for models with and without the number of cigarettes, cigars and pipes, and the number of years of smoking. This agrees with observations in the Swedish case-control study, in which the confounding effect of smoking was adequately addressed by a categorical variable.8 There was also little difference between the effect estimates adjusted for the complete confounder model used in the case-cohort analyses, and adjusted for the limited confounder model used in the full cohort analyses. If the findings from the case-cohort analyses apply to the full cohort, it seems unlikely that residual confounding due to lack of detailed data on smoking and other potential confounders is the explanation of the lack of association in the ex- and current smokers.
A few studies have investigated the association between air pollution and lung cancer incidence in study populations consisting entirely of nonsmokers. A cohort study in the United States (Adventist Health Study on Smog) among 6338 nonsmoking adults found that incident lung cancer was elevated with increased PM10 and SO2 concentrations.6 In a recent case-control study nested within the European Prospective Investigation into Cancer and Nutrition among nonsmokers, lung cancer incidence was possibly increased with residence nearby heavy traffic roads (odds ratio = 1.46 [95% CI = 0.89–2.40]). The odds ratio for NO2 was 1.14 (0.78–1.67) for an increase of 10 μg/m3, and an odds ratio of 1.30 was found for concentrations greater than 30 μg/m3. No clear association was found with PM10 and SO2.7 Our results in the full cohort analyses among never smokers are in agreement with the results among nonsmokers of those other 2 studies.
Some recent European studies investigated the association between air pollution exposure and lung cancer incidence using individual-level exposure assessments, with a focus on traffic-related air pollution. A case-control study in Stockholm on urban air pollution and lung cancer incidence found an RR of 1.05 (95% CI = 0.93–1.18) for an increase of 10 μg/m3 in 30-year average NO2 concentrations from traffic, and a stronger RR of 1.10 (0.97–1.23) when using a 20-year lag (ie, using a 10-year average NO2 concentration over 21–30 years ago).8 In a cohort study among 16,209 men aged 40 to 49 year living in Oslo, the adjusted incidence risk ratio for lung cancer was 1.08 (95% CI = 1.02–1.15) for a 10 μg/m3 increase in average NOx exposure between 1974 and 1978. The risk estimate for a 10 μg/m3 increase in average SO2 exposure between 1974 and 1978 was 1.01 (0.94–1.08).9 These studies suggested an increased risk of lung cancer from traffic-related air pollution, in particular, assessed by either NOx/NO2 concentrations or indicator variables for living near a major road.7–9,25,26 In the study by Nyberg et al,8 heating-related SO2 showed little effect in any time window, despite high exposure levels in the early years of the study period. Although averages of estimated individual SO2 and NO2 exposure showed reasonably high correlation, traffic-related NO2 rather than heating-related SO2 was consistently the stronger risk indicator.
Previous studies showed that the association between lung cancer and air pollution became weaker when more recent exposures were used in the analyses, and that exposures 20 years before the start of the study or even earlier in life were most important.8,9 A limitation of our study is that we estimated exposure to air pollution during the latency period of lung cancer, ie, we estimated exposure based on the baseline 1986 address for 2 periods (1976–1985 and 1987–1996) and we used traffic intensity data for 1986. Correlations between estimated concentrations for different years were however high, even over a period of 20 years (correlation coefficients >0.8).18 It was therefore not possible to evaluate the independent contributions of concentrations of different time periods. It is likely that the correlation with years and periods before 1976 would also be high. We further found that traffic counts from different years over a period of more than 10 years were highly correlated (correlation coefficients >0.9) and we focused mainly on busy roads that have likely been in place for a long period of time. Only the baseline address was available for all subjects, but no complete residential history including exact addresses was available. It was however estimated that 90% of the participants had lived for 10 years or longer in their 1986 municipality. Mean (SD) duration of residence was 35 (19.8) years.19 These results support the use of our exposure estimate as a proxy for time periods further back in time.
A limitation of the exposure assessment method is that we assessed only outdoor concentrations, not taking into account factors related to infiltration of outdoor air pollution into the home, such as air exchange rate. Further, no information was available about the work address. However, approximately 85% of the population had no paid job at baseline. In addition, we had no information about the time participants spent at home or about the time commuting in traffic. The resulting misclassification is, however, likely to be nondifferential.
In a reanalysis of 2 US cohort studies on mortality and long-term exposure to air pollution, higher air pollution effect estimates for lung cancer mortality were found in subjects without high school education.23 We did not find higher effect estimates among subjects with low educational level.
We did find that effect estimates for air pollution and traffic intensity tended to be higher for subjects with low fruit consumption compared with subjects who had medium and high fruit consumption. Such results have not been reported before. Evidence of elevated effect estimates was also found for all-cause mortality among subjects with low fruit consumption.27 Fruit consumption has been consistently inversely associated with lung cancer incidence, with the highest risks for the lowest quintile of fruit consumption.28,29 One explanation is that diet is a source of antioxidants that may protect against oxidative stress.29 It is thought that one of the potential mechanisms of effect of air pollution is also through oxidative stress. Hence, low fruit consumption may be associated with a low protection against oxidative stress effects of air pollution.29
In addition to smoking, occupational exposure to carcinogens, diet and radon exposure are the main determinants of lung cancer. Information about radon exposure was not available in our study. However, only a small percentage of the annual lung cancer cases can be attributed to radon exposure in the Netherlands, and it is principally smokers who are at risk.30 Information about occupational exposure to carcinogens and about diet were available for the case-cohort dataset, but not for the full cohort dataset. Sensitivity analyses showed, however, that results of case-cohort analyses adjusted for the limited confounder model as used in the full cohort analyses are comparable with the case-cohort results adjusted for the complete confounder model including occupational exposure and diet. This suggests that occupational exposure and diet probably did not confound the association between air pollution and lung cancer in our study.
We previously presented results for the association between lung cancer mortality and air pollution in this same population.10 In the full cohort dataset, the adjusted risk estimate for lung cancer mortality for the overall black smoke concentration (period 1987–1996) was 1.03 (0.88–1.20) and the adjusted RR for the traffic intensity on the nearest road was 1.07 (0.96–1.19). Among never smokers, RRs for lung cancer mortality were 1.48 (0.97–2.25) for black smoke overall; 1.15 (0.92–1.46) for the traffic intensity on the nearest road; 1.44 (0.86–2.42) for living near a major road; and 1.36 (0.96–1.93) for the traffic intensity in a 100 m buffer. Effect estimates for lung cancer mortality and lung cancer incidence were thus comparable, which is in agreement with the short time between diagnosis of lung cancer and death from lung cancer (on average 0.9 year [SD 1.0] in our study).
In conclusion, we found evidence for an association between black smoke and lung cancer incidence in people who had never smoked cigarettes. Exposure to traffic was also associated with borderline elevated risks among this group. No clear associations were found for the full cohort or other smoking categories.
The authors are indebted to the participants of this study and further thank the cancer registries and the Netherlands nationwide registry of pathology. They also thank S. van de Crommert, H. Brants, J. Nelissen, C. de Zwart, W. van Dijk, M. Moll, A. Pisters, M. Jansen, H. van Montfort, T. van Moergastel, L. van den Bosch, and R. Schmeitz (Maastricht University/TNO) for assistance, Carla van Wiechen (National Institute of Public Health and the Environment) for advice on geographic information system calculations, and the National Institute of Public Health and the Environment for providing air pollution data monitored by the National Air Quality Monitoring Network in the Netherlands.
1. Laden F, Schwartz J, Speizer FE, et al. Reduction in fine particulate air pollution and mortality: extended follow-up of the Harvard six cities study. Am J Respir Crit Care Med. 2006;173:667–672.
2. Pope CA III, Burnett RT, Thun MJ, et al. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA. 2002;287:1132–1141.
3. Barbone F, Bovenzi M, Cavallieri F, et al. Air pollution and lung cancer in Trieste, Italy. Am J Epidemiol. 1995;141:1161–1169.
4. Jedrychowski W, Becher H, Wahrendorf J, et al. A case-control study of lung cancer with special reference to the effect of air pollution in Poland. J Epidemiol Community Health. 1990;44:114–120.
5. Vineis P, Forastiere F, Hoek G, et al. Outdoor air pollution and lung cancer: recent epidemiologic evidence. Int J Cancer. 2004;111:647–652.
6. Beeson WL, Abbey DE, Knutsen SF. Long-term concentrations of ambient air pollutants and incident lung cancer in California adults: results from the AHSMOG study. Environ Health Perspect. 1998;106:813–823.
7. Vineis P, Hoek G, Krzyzanowski M, et al. Air pollution and risk of lung cancer in a prospective study in Europe. Int J Cancer. 2006;119:169–174.
8. Nyberg F, Gustavsson P, Jarup L, et al. Urban air pollution and lung cancer in Stockholm. Epidemiology. 2000;11:487–495.
9. Nafstad P, Haheim LL, Oftedal B, et al. Lung cancer and air pollution: a 27 year follow up of 16 209 Norwegian men. Thorax. 2003;58:1071–1076.
10. Beelen R, Hoek G, van den Brandt PA, et al. Long-term effects of traffic-related air pollution on mortality in a Dutch cohort. Environ Health Perspect. 2008;116:196–202.
11. Schouten LJ, Höppener P, van den Brandt PA, et al. Completeness of cancer registration in Limburg, The Netherlands. Int J Epidemiol. 1993;22:369–376.
12. Schouten LJ, Jager JJ, van den Brandt PA. Quality of cancer registry data: a comparison of data provided by clinicians with those of registration personnel. Br J Cancer. 1993;68:974–977.
13. van der Sanden GA, Coebergh JW, Schouten LJ, et al. Cancer incidence in The Netherlands in 1989 and 1990: first results of the nationwide Netherlands cancer registry. Coordinating Committee for Regional Cancer Registries. Eur J Cancer. 1995;31A:1822–1829.
14. van den Brandt PA, Goldbohm RA, van't Veer P, et al. A large-scale prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol. 1990;43:285–295.
15. Volovics A, van den Brandt P. Methods for the analyses of case-cohort studies. Biomed J. 1997;39:195–214.
16. van den Brandt PA, Schouten LJ, Goldbohm RA, et al. Development of a record linkage protocol for use in the Dutch cancer registry for epidemiological research. Int J Epidemiol. 1990;19:553–558.
17. Goldbohm RA, van den Brandt PA, Dorant E. Estimation of the coverage of Dutch municipalities by cancer registries and PALGA based on hospital discharge data. Tijdschr Soc Gezond. 1994;72:80–84.
18. Beelen R, Hoek G, Fischer P, et al. Estimated long-term outdoor air pollution concentrations in a cohort study. Atmos Environ. 2007;41:1343–1358.
19. Hoek G, Brunekreef B, Goldbohm S, et al. Association between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet. 2002;360:1203–1209.
20. de Hollander A, Hoeymans N, Melse J, et al. Care for health—Public Health Forecast 2006 (Zorg voor gezondheid. Volksgezondheid Toekomst Verkenning 2006). Houten: Bohn Stafleu Van Loghum; 2006.
21. Lin D, Wei L. The robust inference for the Cox proportional hazards model. J Am Stat Assoc. 1989;84:1074–1078.
22. Sunyer J, Kogevinas M, Kromhout H, et al. Pulmonary ventilatory defects and occupational exposures in a population-based study in Spain. Spanish Group of the European Community Respiratory Health Survey. Am J Respir Crit Care Med. 1998;157:512–517.
23. Krewski D, Burnett R, Goldberg M, et al. Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality. Cambridge: Health Effects Institute; 2000.
24. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7:177–188.
25. Choi KS, Inoue S, Shinozaki R. Air pollution, temperature, and regional differences in lung cancer mortality in Japan. Arch Environ Health. 1997;52:160–168.
26. Tango T. Effect of air pollution on lung cancer: a Poisson regression model based on vital statistics. Environ Health Perspect. 1994;102(Suppl 8):41–45.
27. Smith-Warner SA, Spiegelman D, Yaun SS, et al. Fruits, vegetables and lung cancer: a pooled analysis of cohort studies. Int J Cancer. 2003;107:1001–1011.
28. Linseisen J, Rohrmann S, Miller AB, et al. Fruit and vegetable consumption and lung cancer risk: updated information from the European Prospective Investigation into Cancer and Nutrition (EPIC). Int J Cancer. 2007;121:1103–1114.
29. Kelly FJ. Dietary antioxidants and environmental stress. Proc Nutr Soc. 2004;63:579–585.
30. Health Council of the Netherlands. Radon: Evaluation ‘BEIR VI’. The Hague: Health Council of the Netherlands; 2000.
© 2008 Lippincott Williams & Wilkins, Inc.