International Journal of Epidemiology © International Epidemiological Association 1992
Vol. 21, No. 2 Printed in Great Britain
Occupational and Environmental Hazards Associated with Lung Cancer KARL-HEINZ JOCKEL,* WOLFGANG AHRENS,* HEINZ-ERICH WICHMANIM.** HEIKO BECHER,*t ULRICH BOLM AUDORFF.t INGEBORG JAHN,* BEATE MOLIK,§ EBERHARD GREISER* AND JURGEN TIMMII
Regional comparisons of mortality figures show pronounced differences for lung cancer in urban versus rural areas.1'2 After adjusting for tobacco smoking and occupational hazards most of the differences can be explained by these factors. However, a contribution from ambient air pollution was also identified in some case-control studies. In Erie County (New York) Vena3
classified air pollution into three categories on the basis of total suspended particulates (TSP). He found a crude odds ratio (OR) of 1.56 for people who had lived in highly and intermediately polluted areas for more than 50 years. After adjusting for age, smoking, and occupational exposure this risk was reduced to 1.26 (not significant). An increased risk of lung cancer was also observed in several studies after adjustment for smoking and occupational exposure. Brown et al.* described a 60% increased lung cancer risk for those living close to a zinc smelter in Pennsylvania which emitted arsenic and cadmium compared to a reference area. In Sweden, Pershagen5 found an OR of 2 for people living close to a copper smelter which emitted arsenic, and Jedrychowski et al.6 observed an OR of 1.42 for the highest air pollution exposure level in males in Cracow (Poland). In China a close association was found between lung cancer mortality and indoor burning of 'smoky' coal in regions which showed an increased mortality of lung cancer in females.7
• Bremen Institute for Prevention Research and Social Medicine (BIPS), Griinenstrasse 120, D-2800 Bremen 1, Federal Republic of Germany (FRG). *• University of Wuppertal, Dept. of Labour Safety and Environmental Medicine, GauBstr. 20, D-5600 Wuppertal 1, FRG. t Present address: German Cancer Research Center, Im Neuenheimer Feld 280, D-6900 Heidelberg, FRG. t City of Hamburg, Office of Work, Health and Social Welfare, Dept of Occupational Safety, Division of Occupational Medicine, PO Box 76 01 06, D-2000 Hamburg 76, FRG. § Medical Institute of Environmental Hygiene at the University of Dusseldorf, Auf'm Hennekamp 50, D-4000 Dusseldorf 1, FRG. I University of Bremen, Dept. of Mathematics and Computer Science, Institute of Statistics, PO Box 33 04 40, D-2800 Bremen 33, FRG.
202
Downloaded from http://ije.oxfordjournals.org/ at University of Auckland on July 14, 2015
Jockel K-H (Bremen Institute for Prevention Research and Social Medicine (BIPS), Griinenstrasse 120, D-2800 Bremen 1, FRG), Ahrens W, Wichmann H-E, Becher H, Bolm-Audorff U, Jahn I, Molik B, Greiser E and Timm J. Occupational and environmental hazards associated with lung cancer. International Journal of Epidemiology 1992 21: 202-213. In a hospital-based case-control study 194 lung cancer cases, 194 hospital controls, and 194 population controls were interviewed for their smoking, occupational, and residential history by trained interviewers, using a standardized questionnaire. In order to include many different environmental exposures, case ascertainment took place in seven different hospitals with catchment areas ranging from rural to highly industrialized. Lung cancer risk strongly increases with cumulative cigarette dose, reaching an odds ratio (OR) of 16.19 (95% confidence limits (CD: 5.10, 51.33) for male smokers of more than 40 pack-years and an OR of 19.99 (95% CL: 4.98, 80.24) for female smokers of more than 20 pack-years. For the quantification of occupational exposure to known carcinogens of the lung a novel approach was developed which accumulates exposure information obtained by supplemental questionnaires through an automatic procedure. The OR for the highest exposure group in males was 2.7 (95% CL: 1.23, 5.78). Significantly increased risks were observed in the metal industry, particularly in smelter and foundry workers (OR 4.8, 95% CL: 1.15, 20.16) and in turners (OR 2.2, 95% CL: 1.05, 4.75). In the construction industry the risks were particularly high in road construction workers (OR 3.7, 95% CL: 1.06, 13.20) and in unskilled construction workers (OR 2.7, 95% CL: 1.24, 5.76). The risks in these occupational groups increased with duration and with latency. Quantification of air pollution was done on a county basis by time period. An index based on emission data for sulphur dioxide was compared to a semiquantitative index, which included additional information on ambient air pollution. After adjustment for smoking and occupational exposures an OR of 1.01 (95% CL: 0.53, 1.91) for an emission index and of 1.16 (95% CL: 0.64, 2.13) for a semiquantitative index was obtained.
LUNG CANCER AND OCCUPATIONAL AND ENVIRONMENTAL HAZARDS
MATERIALS AND METHODS Some 194 incident and histologically-confirmed lung cancer cases, 194 hospital and 194 population controls have been included in the study. Cases (146 males, 48 females) were recruited from seven hospitals in five German cities (Bremen, Hamburg, Hannover, Cologne and Bielefeld), which was only a sample of all hospitals in these cities. The case series has thus to be considered as hospital-based. Cities and hospitals were selected to represent a variety of exposures (urban and rural catchment areas) and lung cancer mortality rates (relative differences in rates 50% for males and 80% for females). Cases were defined incident if the diagnosis based on a bronchoscopy was less than 3 months old at the date of the interview. All patients who were suspected of having pulmonary metastases from a different primary tumour were excluded. For 139 (72%) of 194 patients the judgement of a reference pathologist could be obtained. While the classification by histological type showed some discordance," the
diagnosis of primary lung cancer was confirmed for all these cases. The recruitment phase for female cases was extended in order to increase their sample size and allow inferences for both sexes. Hospital-based case-control studies are sometimes criticized because some selection may have occurred. In order to investigate this possible source of bias, a follow-up of all male and female cases which occurred within the recruitment phase in the five cities was performed.12 With regard to the risk factors smoking and occupation, as abstracted from the hospital records, the cases collected for this study were representative for all incident cases. Furthermore, there was no indication of a possible survival bias. Controls were matched to the cases by sex and age (±5 years). In order to avoid overmatching, study centre was not used as a matching variable. Instead, the study centre of a particular control was randomly chosen according to the refinement of a method suggested by Fears and Brown13 which is described in Becher et al.l4 Hospital controls were eligible, if they had an admission diagnosis that was unrelated to smoking. Population controls were randomly drawn from the mandatory residence registries of the above mentioned cities. Of 531 who were eligible as population controls (100%), 216 (40.7%) were interviewed; 49 claimed to be too ill to be interviewed (9.2%), 19 had moved or were deceased (3.6%) and no reply could be obtained from 40 (7.5%). Some 207 (39.0%) were regarded as nonresponders of whom 161 explicitly refused to be interviewed. Some basic characteristics of cases and controls are displayed in Table 1. Data Collection Cases and controls were interviewed by trained interviewers. A structured questionnaire was used to obtain information on job history and occupational exposure, active and passive smoking, residence, dietary habits (vitamin A and vitamin supplements), medical history, leisure time activities, and basic demographic characteristics. All interviewers participated in three 2-day intense central training seminars during the recruitment phase. In addition regular meetings took place in the study centres (every 2 weeks) where all problems arising after visual editing of the questionnaires could be discussed. For a subset of variables from the questionnaire immediate data entry was performed to monitor and reduce possible interviewer effects. Smoking For each smoker (defined as having smoked regularly for more than 6 months) a detailed smoking history
Downloaded from http://ije.oxfordjournals.org/ at University of Auckland on July 14, 2015
Nevertheless, the controversy about the contribution of air pollution to the development of lung cancer persists. The WHO risk assessment group estimated, that in highly polluted areas 5-10 lung cancer cases per 100000 males per year may be attributable to air pollution.8 In parts of the Federal Republic of Germany (FRG) especially in the Ruhr District, air pollution was particularly high until the early 1970s. Benzo(a)pyrene (BaP) concentrations of about 100 ng/m3 were reached, which were higher than in most industrialized areas in Western Europe or the US at that time. Since then, the situation has improved significantly, and the BaP concentrations have decreased to 10 ng/m3 or less since 1980.9 However, for the lung cancer cases of today, exposure of the last 20-40 years has to be considered, due to the long latency period of this tumour. As smoking and exposure to occupational carcinogens are risk factors strongly related to lung cancer, any epidemiological study on the effect of air pollution bears the risk of reporting spurious correlations when an inadequate adjustment for these confounders exists.10 The present case-control study was designed as a pilot study to develop appropriate instruments, to test the feasibility, and to calculate the sample size for a larger study that has the statistical power to detect the small risk elevation for air pollution which was expected to be of the order of 1.1-1.2. Therefore, the number of cases and controls is too small in the present study to allow significant conclusions on the influence of ambient air pollution.
203
204
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 1 Characteristics of cases and controls
Females
Males Cases 3
No.
146
Controls
Cases
Hospital b
Population
146
146
Controls Hospital13
Population
48
48
48
62 38-83
61 38-81
63 38-82
60 40-85
62 38-87
60 40-80
Duration of interview0 Median (minutes) Range (minutes)
60 25-130
55 20-120
65 23-125
50 25-115
50 15-80
60 25-105
Histological cell types Oat cell carcinoma Squamous cell carcinoma Adenocarcinoma Large cell and other
46 62 21 17
Education 0-20 >20-40 >40 Other products only Years since quitting 0-5 >5-10
Use of filter cigarettes Nonfilter cigarettes0
Hospital (No.)
ORa
95% confidence limits
Population (No.)
32
22
43 46 48 6
54 30 23 7
49 45 16 14
7.31 8.33 16.19 4.62
(2.43, 22.23) (2.77, 25.02) (5.10,51.33) (1.15, 18.56)
111 16 16
68 14 32
66 18 40
1 0.90 0.35
(0.44, 1.85) (0.17,0.73)
102 35
94 13
102 8
1 2.44
(1.24,4.79)
Females Nonsmokersb Smokers (pack years) >0-20 >20
14 24
15 2
19 8
5.72 19.99
(1.32, 24.74) (4.98, 80.24)
Years since quitting 0-5 >5
35 3
12 5
17 10
1 0.18
(0.03, 1.12)
a b c
21
10
Both control groups combined in the analysis (1:2 matched design). Never smokers or persons smoking regularly for less than half a year. Within the last 20 years (1965-1985).
Downloaded from http://ije.oxfordjournals.org/ at University of Auckland on July 14, 2015
since the beginning of the century. All counties were classified into eight pollution categories for 10-year intervals from 1895 to 1984.
207
LUNG CANCER AND OCCUPATIONAL AND ENVIRONMENTAL HAZARDS
Occupation Males. The mean duration of lifetime employment was between 36.5 and 37.4 years for cases and controls, respectively. The average number of jobs ranged from 6.4 (hospital controls) to 6.8 (population controls), with cases falling between. Table 4 shows the number of male cases and controls and OR by major branch of industry. Each subject could be counted for more than one job. The average count in different industries was 2.8 for both cases and population controls, and 2.5 for hospital controls, indicating that there was no differential underreporting of jobs. The risk estimates exceeded the value of 1.0 significantly (P
Vol. 21, No. 2 Printed in Great Britain
Occupational and Environmental Hazards Associated with Lung Cancer KARL-HEINZ JOCKEL,* WOLFGANG AHRENS,* HEINZ-ERICH WICHMANIM.** HEIKO BECHER,*t ULRICH BOLM AUDORFF.t INGEBORG JAHN,* BEATE MOLIK,§ EBERHARD GREISER* AND JURGEN TIMMII
Regional comparisons of mortality figures show pronounced differences for lung cancer in urban versus rural areas.1'2 After adjusting for tobacco smoking and occupational hazards most of the differences can be explained by these factors. However, a contribution from ambient air pollution was also identified in some case-control studies. In Erie County (New York) Vena3
classified air pollution into three categories on the basis of total suspended particulates (TSP). He found a crude odds ratio (OR) of 1.56 for people who had lived in highly and intermediately polluted areas for more than 50 years. After adjusting for age, smoking, and occupational exposure this risk was reduced to 1.26 (not significant). An increased risk of lung cancer was also observed in several studies after adjustment for smoking and occupational exposure. Brown et al.* described a 60% increased lung cancer risk for those living close to a zinc smelter in Pennsylvania which emitted arsenic and cadmium compared to a reference area. In Sweden, Pershagen5 found an OR of 2 for people living close to a copper smelter which emitted arsenic, and Jedrychowski et al.6 observed an OR of 1.42 for the highest air pollution exposure level in males in Cracow (Poland). In China a close association was found between lung cancer mortality and indoor burning of 'smoky' coal in regions which showed an increased mortality of lung cancer in females.7
• Bremen Institute for Prevention Research and Social Medicine (BIPS), Griinenstrasse 120, D-2800 Bremen 1, Federal Republic of Germany (FRG). *• University of Wuppertal, Dept. of Labour Safety and Environmental Medicine, GauBstr. 20, D-5600 Wuppertal 1, FRG. t Present address: German Cancer Research Center, Im Neuenheimer Feld 280, D-6900 Heidelberg, FRG. t City of Hamburg, Office of Work, Health and Social Welfare, Dept of Occupational Safety, Division of Occupational Medicine, PO Box 76 01 06, D-2000 Hamburg 76, FRG. § Medical Institute of Environmental Hygiene at the University of Dusseldorf, Auf'm Hennekamp 50, D-4000 Dusseldorf 1, FRG. I University of Bremen, Dept. of Mathematics and Computer Science, Institute of Statistics, PO Box 33 04 40, D-2800 Bremen 33, FRG.
202
Downloaded from http://ije.oxfordjournals.org/ at University of Auckland on July 14, 2015
Jockel K-H (Bremen Institute for Prevention Research and Social Medicine (BIPS), Griinenstrasse 120, D-2800 Bremen 1, FRG), Ahrens W, Wichmann H-E, Becher H, Bolm-Audorff U, Jahn I, Molik B, Greiser E and Timm J. Occupational and environmental hazards associated with lung cancer. International Journal of Epidemiology 1992 21: 202-213. In a hospital-based case-control study 194 lung cancer cases, 194 hospital controls, and 194 population controls were interviewed for their smoking, occupational, and residential history by trained interviewers, using a standardized questionnaire. In order to include many different environmental exposures, case ascertainment took place in seven different hospitals with catchment areas ranging from rural to highly industrialized. Lung cancer risk strongly increases with cumulative cigarette dose, reaching an odds ratio (OR) of 16.19 (95% confidence limits (CD: 5.10, 51.33) for male smokers of more than 40 pack-years and an OR of 19.99 (95% CL: 4.98, 80.24) for female smokers of more than 20 pack-years. For the quantification of occupational exposure to known carcinogens of the lung a novel approach was developed which accumulates exposure information obtained by supplemental questionnaires through an automatic procedure. The OR for the highest exposure group in males was 2.7 (95% CL: 1.23, 5.78). Significantly increased risks were observed in the metal industry, particularly in smelter and foundry workers (OR 4.8, 95% CL: 1.15, 20.16) and in turners (OR 2.2, 95% CL: 1.05, 4.75). In the construction industry the risks were particularly high in road construction workers (OR 3.7, 95% CL: 1.06, 13.20) and in unskilled construction workers (OR 2.7, 95% CL: 1.24, 5.76). The risks in these occupational groups increased with duration and with latency. Quantification of air pollution was done on a county basis by time period. An index based on emission data for sulphur dioxide was compared to a semiquantitative index, which included additional information on ambient air pollution. After adjustment for smoking and occupational exposures an OR of 1.01 (95% CL: 0.53, 1.91) for an emission index and of 1.16 (95% CL: 0.64, 2.13) for a semiquantitative index was obtained.
LUNG CANCER AND OCCUPATIONAL AND ENVIRONMENTAL HAZARDS
MATERIALS AND METHODS Some 194 incident and histologically-confirmed lung cancer cases, 194 hospital and 194 population controls have been included in the study. Cases (146 males, 48 females) were recruited from seven hospitals in five German cities (Bremen, Hamburg, Hannover, Cologne and Bielefeld), which was only a sample of all hospitals in these cities. The case series has thus to be considered as hospital-based. Cities and hospitals were selected to represent a variety of exposures (urban and rural catchment areas) and lung cancer mortality rates (relative differences in rates 50% for males and 80% for females). Cases were defined incident if the diagnosis based on a bronchoscopy was less than 3 months old at the date of the interview. All patients who were suspected of having pulmonary metastases from a different primary tumour were excluded. For 139 (72%) of 194 patients the judgement of a reference pathologist could be obtained. While the classification by histological type showed some discordance," the
diagnosis of primary lung cancer was confirmed for all these cases. The recruitment phase for female cases was extended in order to increase their sample size and allow inferences for both sexes. Hospital-based case-control studies are sometimes criticized because some selection may have occurred. In order to investigate this possible source of bias, a follow-up of all male and female cases which occurred within the recruitment phase in the five cities was performed.12 With regard to the risk factors smoking and occupation, as abstracted from the hospital records, the cases collected for this study were representative for all incident cases. Furthermore, there was no indication of a possible survival bias. Controls were matched to the cases by sex and age (±5 years). In order to avoid overmatching, study centre was not used as a matching variable. Instead, the study centre of a particular control was randomly chosen according to the refinement of a method suggested by Fears and Brown13 which is described in Becher et al.l4 Hospital controls were eligible, if they had an admission diagnosis that was unrelated to smoking. Population controls were randomly drawn from the mandatory residence registries of the above mentioned cities. Of 531 who were eligible as population controls (100%), 216 (40.7%) were interviewed; 49 claimed to be too ill to be interviewed (9.2%), 19 had moved or were deceased (3.6%) and no reply could be obtained from 40 (7.5%). Some 207 (39.0%) were regarded as nonresponders of whom 161 explicitly refused to be interviewed. Some basic characteristics of cases and controls are displayed in Table 1. Data Collection Cases and controls were interviewed by trained interviewers. A structured questionnaire was used to obtain information on job history and occupational exposure, active and passive smoking, residence, dietary habits (vitamin A and vitamin supplements), medical history, leisure time activities, and basic demographic characteristics. All interviewers participated in three 2-day intense central training seminars during the recruitment phase. In addition regular meetings took place in the study centres (every 2 weeks) where all problems arising after visual editing of the questionnaires could be discussed. For a subset of variables from the questionnaire immediate data entry was performed to monitor and reduce possible interviewer effects. Smoking For each smoker (defined as having smoked regularly for more than 6 months) a detailed smoking history
Downloaded from http://ije.oxfordjournals.org/ at University of Auckland on July 14, 2015
Nevertheless, the controversy about the contribution of air pollution to the development of lung cancer persists. The WHO risk assessment group estimated, that in highly polluted areas 5-10 lung cancer cases per 100000 males per year may be attributable to air pollution.8 In parts of the Federal Republic of Germany (FRG) especially in the Ruhr District, air pollution was particularly high until the early 1970s. Benzo(a)pyrene (BaP) concentrations of about 100 ng/m3 were reached, which were higher than in most industrialized areas in Western Europe or the US at that time. Since then, the situation has improved significantly, and the BaP concentrations have decreased to 10 ng/m3 or less since 1980.9 However, for the lung cancer cases of today, exposure of the last 20-40 years has to be considered, due to the long latency period of this tumour. As smoking and exposure to occupational carcinogens are risk factors strongly related to lung cancer, any epidemiological study on the effect of air pollution bears the risk of reporting spurious correlations when an inadequate adjustment for these confounders exists.10 The present case-control study was designed as a pilot study to develop appropriate instruments, to test the feasibility, and to calculate the sample size for a larger study that has the statistical power to detect the small risk elevation for air pollution which was expected to be of the order of 1.1-1.2. Therefore, the number of cases and controls is too small in the present study to allow significant conclusions on the influence of ambient air pollution.
203
204
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 1 Characteristics of cases and controls
Females
Males Cases 3
No.
146
Controls
Cases
Hospital b
Population
146
146
Controls Hospital13
Population
48
48
48
62 38-83
61 38-81
63 38-82
60 40-85
62 38-87
60 40-80
Duration of interview0 Median (minutes) Range (minutes)
60 25-130
55 20-120
65 23-125
50 25-115
50 15-80
60 25-105
Histological cell types Oat cell carcinoma Squamous cell carcinoma Adenocarcinoma Large cell and other
46 62 21 17
Education 0-20 >20-40 >40 Other products only Years since quitting 0-5 >5-10
Use of filter cigarettes Nonfilter cigarettes0
Hospital (No.)
ORa
95% confidence limits
Population (No.)
32
22
43 46 48 6
54 30 23 7
49 45 16 14
7.31 8.33 16.19 4.62
(2.43, 22.23) (2.77, 25.02) (5.10,51.33) (1.15, 18.56)
111 16 16
68 14 32
66 18 40
1 0.90 0.35
(0.44, 1.85) (0.17,0.73)
102 35
94 13
102 8
1 2.44
(1.24,4.79)
Females Nonsmokersb Smokers (pack years) >0-20 >20
14 24
15 2
19 8
5.72 19.99
(1.32, 24.74) (4.98, 80.24)
Years since quitting 0-5 >5
35 3
12 5
17 10
1 0.18
(0.03, 1.12)
a b c
21
10
Both control groups combined in the analysis (1:2 matched design). Never smokers or persons smoking regularly for less than half a year. Within the last 20 years (1965-1985).
Downloaded from http://ije.oxfordjournals.org/ at University of Auckland on July 14, 2015
since the beginning of the century. All counties were classified into eight pollution categories for 10-year intervals from 1895 to 1984.
207
LUNG CANCER AND OCCUPATIONAL AND ENVIRONMENTAL HAZARDS
Occupation Males. The mean duration of lifetime employment was between 36.5 and 37.4 years for cases and controls, respectively. The average number of jobs ranged from 6.4 (hospital controls) to 6.8 (population controls), with cases falling between. Table 4 shows the number of male cases and controls and OR by major branch of industry. Each subject could be counted for more than one job. The average count in different industries was 2.8 for both cases and population controls, and 2.5 for hospital controls, indicating that there was no differential underreporting of jobs. The risk estimates exceeded the value of 1.0 significantly (P
