Inhalation of radon is the main source of exposure to radioactivity in the general population of most countries. 1 Radon concentrations can be markedly increased indoors because air space is confined, exchange with outside air is limited, and lower air pressure in buildings favors penetration of radon from the soil. 2 Exhalation from radium-rich building materials, used in about 10% of the Swedish housing stock, and release of radon dissolved in water can also in some circumstances contribute to indoor radon levels. Radon gas decays into short-lived products emitting high-energy alpha particles that irradiate the bronchial epithelium after inhalation and deposition. 3
Numerous epidemiologic studies have been conducted to quantify the association between lung cancer and radon exposure both in occupational settings, such as among miners, and in the general population. Epidemiologic studies of residential radon provide effect estimates based on levels experienced by the general population and thereby avoid the extrapolation from high to low exposure levels that occurs when effect estimates are obtained from miner studies. Nevertheless, the overall patterns of risk ratios in the two settings show good agreement and support an excess lung cancer risk with increased levels of radon exposure. 4–6 Experimental animal studies show that interaction with other environmental agents is not necessary for the induction of lung cancer by radon, but the experimental exposure rates are not commensurable with average levels in dwellings. 7
An important issue, which requires further clarification, concerns the risk posed by radon for never-smokers. The miner data suggest a steeper increase in relative risk with radon exposure among never-smokers than among smokers. 4 By contrast, the summary exposure-response estimates from meta-analyses of residential radon studies appear similar with and without restriction to subjects who never smoked, respectively, although the effect estimate among never-smokers is uncertain owing to few lung cancer cases. 5 Results among miners have been more influential for risk assessment than those conducted in the general public, although inference about never-smokers is also limited, as it relies on a total of only 64 cases of lung cancer in miner studies. 4,8 The present work aims to strengthen the evidence regarding the radon-related lung cancer effect among never-smokers in the general population. In addition, we investigate for the first time the interaction between exposure to environmental tobacco smoke (ETS) and residential radon in relation to lung cancer.
Subjects and Methods
The database for never-smokers in five case-control studies (individually described below) was complemented with measurements of radon concentrations in dwellings occupied during the retrospective 32-year period ending 3 years before assessment of disease status (hereafter referred to as the observation period for radon exposure). These studies were originally conducted for purposes of evaluating the association of lung cancer with environmental and occupational exposures other than radon. The subjects from these studies were analyzed here with respect to radon exposure for the first time and were combined for additional analyses with the never-smokers from a previous study of residential radon exposure and lung cancer in Sweden. All never-smoker case subjects in participating studies were included provided that their year of diagnosis was 1980 or later. A subject who had not smoked daily for 1 year or longer was defined as a never-smoker. We selected up to twice as many never-smoker control subjects, matched on birth year (up to a 3-year difference), sex, and study, as case subjects. The date of diagnosis for a case determined the end of follow-up for the case subject and his or her matched controls. For some controls, their own date of death marked their end of follow-up, if death occurred before (up to 3 years) the diagnosis of the matched case. These instances occurred for Study 2 (below), in which controls had been originally frequency-matched to the cases on age at calendar year of diagnosis, using 5-year age groups, and for Study 3, in which controls had been originally frequency-matched to the cases on time period for occurrence of death. Such corrections for the date of end of follow-up among controls were also necessary among the other studies but only on the level of months.
We obtained the addresses for the residences occupied by study subjects for at least 2 years during the observation period for radon exposure through parish registers, as a complement to the residential history that was also obtained from questionnaires in most of the participating studies. Measurements of radon concentrations were performed during the heating season, that is, between October and May, in 1996/1997 and 1997/1998, using two detectors per dwelling placed for 3 months in the bedroom and the living room. The detectors were placed and collected by qualified public health officers and were processed at the Swedish Radiation Protection Institute. A detector consists of a holder containing an alpha-track detector material, CR-39 (allyl diglycol carbonate), and the track density of the film is measured by an automatic film reader after chemical etching. 9 For each address, the radon concentration was estimated by the average of the measured radon levels in bedroom and living room.
The length of the observation period for radon exposure was 32 years, except for 11 subjects younger than 35 years of age at the end of follow-up. The time-weighted average (TWA) radon concentration for a study subject during the observation period was based on radon concentrations over all addresses occupied by the subject during that time period, weighting concentrations proportionally to the number of years spent at each address. For the time intervals for which no measurement was available, we imputed the study-specific arithmetic average of radon concentrations for measured addresses of control subjects. 10 For the supplementary data from the nationwide Swedish study described below, averages of measured addresses among controls were estimated for three groups of municipalities classified according to radon risk and used for imputation of unmeasured addresses within corresponding groups. 10
Other Risk Factors
Unless data had already been collected by the contributing study, information on ETS exposure was obtained through telephone interview of study subjects or proxies, using a questionnaire containing questions about domestic exposure to ETS during adulthood, including intensities and time periods of smoking by spouse or cohabitant. Subjects were classified into one of four categories according to whether they had ever been exposed to ETS at home and whether the lapse of time since last exposure to ETS was equal to or larger than 15 years, between 15 and 3 years, or less than 3 years. Such classification appeared to be more appropriate for relating the excess risk of lung cancer to ETS exposure than the simple dichotomy ever-/never-exposed. 11
Two dichotomous variables for occupational lung cancer risk were constructed. One variable distinguished subjects who, for a year or more during their whole occupational history, had held occupations recognized to present lung cancer excess risk, whereas the other variable distinguished subjects who had held occupations suspected to present lung cancer excess risk. 12,13 Classification relied on Scandinavian occupational codes (NYK-83) 14 and, for Study 1 (below), on the cross-classification of international occupational (ISCO-68) and industrial codes (ISIC-71). 13
We constructed a socioeconomic indicator variable with four levels by combining two dichotomies indicating whether, on a TWA basis, the jobs reported in the working history predominantly required short or long education and manual work or not. 15
We constructed a dichotomous variable for urbanization by distinguishing subjects who lived for 10 years or more in one of the three largest cities in Sweden (Stockholm, Göteborg, or Malmö) at some time during the follow-up period.
Summary of Data Sources
The relation between ETS exposure and lung cancer was investigated among never-smokers, 30 years of age or older, residing in Stockholm County. 11 Cases were recruited between October 1, 1989, and September 30, 1995, among patients referred to the three main hospitals responsible for medical investigation and treatment of lung cancer in this area. All suspected lung cancer cases were screened for smoking status, and all never-smokers were traced until they either were included in the study when diagnosed as having lung cancer [International Classification of Diseases (ICD), 9th revision, code 162] or were excluded. About twice as many controls were frequency-matched to the case subjects by age group (30–49, 50–69, and 70 years and older), sex, and hospital catchment area. Interviews were conducted in person or by telephone using a detailed questionnaire on exposure to ETS, occupational and residential histories, and other relevant exposures. Answers regarding never-smoking status and ETS were validated through a secondary interview with a next-of-kin. 16 The participation rate was 85.5% among cases and 82.9% among controls screened for never-smoking status (6.1% of controls potentially eligible as never-smokers could not be contacted even for screening). All 124 cases and 235 controls were selected for the present study. The basis for diagnosis was tissue histology for 79.0% and cytology for 20.2% of the case subjects.
With the primary aim to clarify the relation between environmental or occupational air pollution and lung cancer, a case-control study was conducted among men 40 to 75 years of age and resident in Stockholm County at some time between January 1, 1985, and December 31, 1990, and with no more than 5 years of residence outside of the county between 1950 and 1990. 15,17 A total of 1,196 eligible cases were identified, using the Stockholm County regional cancer registry, among cases of lung cancer (ICD-7, code 162.1) diagnosed between January 1, 1985, and December 31, 1990. About twice as many control subjects, frequency-matched to the cases on age at calendar year of diagnosis (in 5-years age groups), were randomly selected in 1992 from population registers of Stockholm County for the corresponding years. About half of the controls were additionally frequency-matched to the cases on vital status by 1990, using the Cause-of-Death Registry, excluding those with a smoking-related cause of death. A questionnaire was mailed to subjects or their next-of-kin, including questions about smoking, ETS exposure at home, and occupational and residential history, and was followed by phone contact. Among the 1,196 cases, and the 1,324 and 1,441 eligible controls matched and not matched for vital status, the participation rates were 87%, 82%, and 88%, respectively. All 33 never-smokers with lung cancer and twice as many controls were selected for the present study. The basis for diagnosis among never-smoker cases was tissue histology or cytology in 81.8% and 18.2%, respectively.
A case-control study designed to assess the excess risk of lung cancer related to environmental emissions from Rönnskärsverken, a large smelting plant in northern Sweden, recruited cases and control subjects among men and women who had died in Skellefteå Municipality between 1961 and 1990, excluding those who had worked at Rönnskärsverken or as miners. 18 Subjects with a diagnosis of lung cancer (ICD-7, code 162) were identified using the Regional Cancer Registry and the National Cause-of-Death Registry. Twice as many controls, matched to cases on birth year and sex, were selected from the National Cause-of-Death Registry. A questionnaire about smoking habits, occupations, and residential history was mailed to next-of-kin and was often complemented by telephone interviews. The response rate for the 378 lung cancer cases identified was 94%, and about 93% for controls. For the present study, 78 never-smoking control subjects and 40 never-smoking cases diagnosed in 1980 or later were selected. The basis for diagnosis for these never-smoking cases was tissue histology or cytology for 70.0% and 30.0%, respectively.
A case-control study, undertaken to investigate the relation between environmental factors and lung cancer, was conducted among subjects less than 75 years of age and living at the time of the study in one of 26 municipalities in Gothenburg and Bohus County and Älvsborg County, in southwest Sweden. 19 The municipalities were selected to represent the referral area of the three main regional hospitals where cases were recruited between January 1989 and June 1994, with occasional interruptions, among patients with suspected lung cancer referred to the pulmonary units. Potential cases and controls were matched for county, sex, and birth date by selecting controls registered on date of birth next to the cases within population registers. Interviews using a questionnaire to assess smoking, ETS, occupational exposures, and more were conducted in person, with a participation rate of 86% among cases and 75% among controls. Only suspected cases later diagnosed and recorded in the Regional Cancer Registry as having primary lung cancer (ICD-7, code 162.1) were retained (52%). Of the resulting 539 cases, all 49 never-smoker cases and twice as many never-smoker controls were selected for the present study. The basis for diagnosis was tissue histology or cytology for 87.8% and 10.2% of the cases, respectively.
A case-control study was designed to investigate the association between occupational exposure and lung cancer. The study was conducted in a Swedish county, Västernorrland, containing a large number of paper and pulp mills. Cases were men diagnosed with primary lung cancer (ICD-7, code 162.1) between 1978 and 1991 reported to the Regional Cancer Registry in Umeå and deceased before September 1, 1992. Deceased controls were selected from the National Cause-of-Death Registry and matched to cases on year of birth, year of death, and municipality of residence at the time of diagnosis of the matched case. Subjects with suicide and lung cancer as cause of death were not accepted as controls. For each case not over 80 years of age at the time of the investigation, an additional living control, matched on year of birth and municipality of residence at the time of diagnosis of the case, was selected from the national population register. Information on occupational history and smoking habits was obtained through a questionnaire mailed to the study subjects or their next-of-kin. Participation rates were 88% among cases and 84% among controls. Among the resulting 562 cases and 878 controls, all 25 never-smoker cases, diagnosed in 1980 or later, and twice as many controls were included in the present study. The basis for diagnosis for the never-smoker cases was tissue histology or cytology for 76% and 24%, respectively.
Complementary analyses were performed using data from the nationwide Swedish case-control study for which the radon measurement procedures were similar. 20 This study included 1,360 cases of primary lung cancer (ICD-7, code 162.1) 35–74 years of age, diagnosed from 1980 to 1984, and 2,847 population controls. For eligibility, subjects had to live in Sweden on January 1, 1947, and had to reside at the time of selection in 1 of 109 municipalities that were regarded likely to be either low- or high-radon areas. In one group of controls, subjects sampled from the study base were frequency-matched for age (5-year intervals) and calendar year to the lung cancer cases. In another group, control subjects were additionally matched on vital status to the cases, excluding smoking-related causes of death. Radon was measured in 8,992 dwellings, and a measurement could not be obtained in 3,402 dwellings (27.4%), usually because the house no longer existed or was now being used only during summer. Information on potential risk factors for lung cancer was obtained from study subjects or next-of-kin through questionnaires supplemented by telephone interviews. The response rate was 82.2% for cases and 81.7% for controls. After selection of all never-smokers with at least one radon measurement within the 32-year observation period, data on 178 cases and 1,162 controls were available. The basis for diagnosis was tissue histology or cytology for 80.3% and 18.5% of the never-smoking cases, respectively.
The data were analyzed using the Stata and Epicure statistical packages. 21,22 We estimated the excess relative risk (ERR) of lung cancer per unit of TWA residential radon concentration in conditional logistic regression analyses using a continuous variable for the TWA radon concentrations in dwellings of individual subjects. The 95% confidence limits for the ERR were based on the likelihood-ratio criterion. 23,24 A linear relative risk (RR) function was fitted to the data, RR = 1 + βx, where β is the ERR parameter and x denotes TWA residential radon level. 25,26 Maximum likelihood estimates of relative risks for categories of TWA residential radon level were obtained for categories delimited by cutpoints used in the nationwide Swedish study. 20 The indicator variable used for building the conditioning strata in the logistic regression model represented study, age group (5-year intervals), sex, and matched area of current residence, and to adjust further the effect of radon exposure, we added as predictors in the model indicator variables for ETS exposure, urbanization, occupational lung cancer risk, and socioeconomic status. Building these conditioning strata involved post hoc stratification because the original matching had sometimes matched never-smoking case subjects with smoking control subjects. For these subjects the matching was broken and reconstructed with never-smoking controls with corresponding levels of the matched factors. Given similar levels of the matching factors, the information on which controls and cases had been matched originally or in a second stage is irrelevant for the purpose of matching. In addition, individually matched pairs can be split to perform more valid analyses that condition on strata grouping of subjects with similar characteristics rather than on matched pairs, 27 and it is also appropriate to include factors not matched in the design stage within the conditioning strata for logistic regression, which allows combined analysis of studies with different matching factors. 28
On the average, about 25 years of a subject’s 32-year residential history was covered by measurements. There were 2,052 residential periods of at least 2 years during the observation period for radon exposure, and radon measurements at corresponding addresses were obtained for 1,595 residential periods (77.7%). The rates for achieving radon measurements were independent of disease status, and this also applies to the reasons for not obtaining measurements (Table 1).
Of the 798 study subjects enrolled (271 lung cancer cases and 527 controls), 745 subjects (258 cases and 487 controls) with at least one address assessed for radon level were included in risk analyses (Table 2). The distribution of these 745 study subjects over number of addresses occupied, and over number of addresses measured, was comparable for cases and controls. The mean period of residence for which direct measurements were available was 26.1 years among case subjects and 26.8 years among control subjects, or 81.7% and 83.8% of the defined exposure period, respectively. Information about risk factors other than radon, obtained through questionnaires, was provided by the subjects themselves (rather than by next-of-kin) for 65.5% of the case subjects and for 76.0% of the controls.
The distributions of cases and controls over the potential confounders, presented in Table 3, suggest differences in risk of lung cancer according to ETS exposure, occupational exposure, and socioeconomic classification. This finding was confirmed in logistic regression analyses stratified for matched factors (age, sex, study center, and matched area of residence) and adjusted for the other risk factors. Except for matching factors, adjustment had very little influence on risk estimates for radon exposure.
The distribution of TWA radon concentrations among cases was shifted toward higher levels compared with controls, as can be assessed by comparing the group-specific averages and quartiles in Table 4. The contrast between case and control groups with respect to the distribution of TWA radon concentrations was unaffected when gaps in reconstructing the exposure history were imputed with average radon concentrations, although this procedure reduced the variability of the distribution.
We found a trend in the relative risks of lung cancer with increasing radon exposure categories in the current study that is much like the trend shown by the relative risks among never-smokers in the nationwide Swedish study (Table 5). The ERR based on a continuous variable for radon appeared larger in the current study than in the nationwide study. The range of exposure was larger in that study (maximum TWA radon exposure = 2,554 Bq m− 3), and there were relatively more controls than cases in the upper range of TWA radon exposures.
Additional analyses evaluated the interaction between residential radon exposure and ETS at home (Table 6). The trend of increased risk with increasing radon levels seems to be limited to subjects exposed to ETS at home, and this pattern was consistent both in the current study as well as among never-smokers in the nationwide study.
Exploratory analyses using only subjects with a large proportion of their observation period for radon exposure covered by actual measurements, or comparing fixed- and random-effect models, did not indicate sensitivity of the present results to these issues. We investigated sensitivity to outliers by restricting the relatively wider range of exposure in the data from the nationwide Swedish study to the range observed in the current study (662 Bq m− 3). This analysis changed the ERR per 100 Bq m− 3 from 0.04 to 0.14, which is closer to the ERR of 0.28 per 100 Bq m− 3 observed in the current study. For the combined data, the ERR was equal to 0.19 per 100 Bq m− 3 when the range of exposure was restricted, instead of the ERR of 0.10 per 100 Bq m− 3 when the range was unrestricted. To get an indication as to whether these subjects might be appropriately omitted from regression analyses, we inspected the radon concentrations indicated by the two measuring devices used in each of the 30 dwellings occupied by the 12 study subjects with TWA exposure greater than 662 Bq m− 3, but each pair of measurements showed good agreement.
The goal of the radon measurement program was to perform measurements in each dwelling occupied by study subjects for at least 2 years during the past 3–35 years. For each single substudy, including never-smokers in the nationwide Swedish study, radon measurements covered at least 78% of the exposure window, which compares favorably with those residential radon studies for which the exposure window covers about three decades. Although lack of measurement for some dwellings cannot be avoided, it is important that the failure rates should be similar for the case group and the control group, because lack of measurement could be related to radon levels. Furthermore, bias would be more plausible if reasons for lack of measurement differed between the two groups of subjects. None of this was apparent in the present study, either when dwellings were taken as a unit for assessment or when the units were the subjects.
The radon levels observed in the new data are lower than in the never-smoker data from the nationwide Swedish study in which the arithmetic mean for TWA radon concentrations was 108 Bq m− 3 (111 Bq m− 3 for cases, 107 Bq m− 3 for controls). The percentages of TWA radon concentrations that exceeded the levels of 200 and 400 Bq m− 3 in these two datasets were 4.8% and 0.8% for the current data and 9.6% and 2.0% for never-smokers in the nationwide Swedish study. In that study, subjects were selected from municipalities likely to be either low- or high-radon areas, which explains the greater representation of higher radon exposure levels. The relative risks obtained in the present reanalysis were slightly different from those previously published, as the imputation method was different and the length of the exposure window slightly shorter, 20 which also explains why there were two subjects from that study who had no dwelling measured during the observation period for radon exposure in the present analysis.
The ERR of 0.10 per 100 Bq m− 3 observed in the present study for all subjects combined is consistent with the ERR of 0.12 at 100 Bq m− 3 TWA radon concentration (95% confidence limits of about −0.15 and 0.4) produced in a previous meta-analysis of nonsmokers from five residential studies that included 777 never-smoker and 161 longtime former-smoker cases (the ERR is also consistent with the overall results for smokers and nonsmokers combined from pooled residential studies or pooled miner studies). 4,5,8,20,29–32 The ERR from the meta-analysis of nonsmokers was mostly driven by the results from a single study. 5
Miner studies have reported a positive interaction between radon and smoking that is less than multiplicative, because the radon-related ERR was greater for never-smokers than for smokers. 4,8,33 The study of interactions generally requires more statistical information than main analyses, and although miner studies benefit from a wide range of exposure, there were few cases of lung cancer (a total of 64) among never-smokers. 4,8 The ten epidemiologic studies of residential radon conducted so far in the general population, with a minimum of 200 subjects and long-term radon measurements in at least one of the subjects’ dwellings, include a total of about 900 never-smoker cases of lung cancer. 20,29–32,34–38 In contrast to the findings among miners, the ERRs among smokers tend to be larger than among never-smokers, in residential radon studies in which such comparison is possible, 20,29,31,32,34,35,38 with possibly an exception for a study in Stockholm, Sweden, in which interpretation is complicated by sensitivity of the results to the choice of a categorical or continuous radon exposure variable for analyses. 30,39 An assessment of heterogeneity of ERR by ETS status was done here for the first time in a radon study, and an interaction that is at least multiplicative between ETS and radon was suggested. As discussed above, this would be in line with the type of interaction observed between radon and smoking in residential studies, but it may be specific to residential radon levels and not applicable over the entire range of exposure found in miner studies.
Contemporary measurements of radon concentrations were used to represent concentrations up to four decades ago. Preceding work suggests that the ERR estimates are likely to be underestimated owing to random error in exposure estimates; in Swedish studies underestimation by a factor of nearly 2 is probable, 40,41 and a comparably large impact on the risk estimates has subsequently been shown to hold for a residential radon study in south-west England. 34 Additional bias could occur if 3-month measurements during the heating season overestimated yearly concentrations, although preliminary results from work in progress show that the effects are probably small.
Omitting subjects with the largest amount of imputation from analysis had little consequence on the ERR estimate, as most subjects had a very large proportion of their observation period for radon exposure covered by actual measurements. By contrast, excluding subjects with large radon exposure values tended to increase the ERR. The main contributor to random error in exposure estimation is the variation in radon concentrations over calendar years in the same dwelling, 41 and therefore the agreement verified between the readings from the two detectors in each dwelling of these subjects, which rules out a measurement error, does not preclude imprecision in exposure assessment. Because large estimated exposures on the average correspond to smaller true exposures and induce downward curvature in a true linear exposure response, 41 an increase in the ERR estimate may be expected after exclusion of such subjects.
In conclusion, the present study adds to the limited data on radon exposure and lung cancer in never-smokers. Overall, the radon-related ERR of lung cancer among never-smokers does not appear to depart much from that in smokers (but is obviously much smaller in terms of absolute excess risk). Among never-smokers, residential radon exposure may be more harmful for those exposed to ETS.
We thank Katinka Almrén of the Department of Environmental Medicine, Karolinska Institutet; Hillevi Giertz of the Swedish Radiation Protection Institute and the staff of the municipal public health boards for assistance with the radon measurements; and Helena Svensson of the Department of Environmental Medicine for skillful data management and programming.
1. World Health Organization. Air Quality Guidelines for Europe. European Series No. 23. Copenhagen: WHO Regional Office for Europe, 1987.
2. Swedjemark GA. Radon and its decay products in housing (Doctoral thesis). Stockholm, Sweden: University of Stockholm, 1985.
3. International Commission on Radiological Protection (ICRP). Lung cancer risk from indoor exposures to radon daughters. Ann ICRP 1987; 17:1–60.
4. National Research Council, Committee on Biological Effects of Ionizing Radiation (BEIR VI). Health Effects of Exposure to Radon. Washington DC: National Academy Press, 1999.
5. Lubin JH, Boice JD Jr. Lung cancer risk from residential radon: meta-analysis of eight epidemiologic studies. J Natl Cancer Inst 1997; 89:49–57.
6. Lubin JH, Boice JD Jr, Edling C, Hornung RW, Howe G, Kunz E, Kuziak RA, Morrison HI, Radford EP, Samet JM, Tirmarche M, Woodward A, Yao SX, Pierce DA. Lung Cancer and Radon: A Joint Analysis of 11 Underground Miner Studies. National Cancer Institute. NIH Pub. No. 94-3644. Washington DC: U.S. Department of Health and Human Services, 1994.
7. Cross FT. Residential radon risk from the perspective of experimental animal studies. Am J Epidemiol 1994; 140:333–339.
8. Lubin JH, Boice JD Jr, Edling C, Hornung RW, Howe G, Kunz E, Kuziak RA, Morrison HI, Radford EP, Samet JM, Tirmarche M, Woodward A, Yao SX, Pierce DA. Lung cancer in radon-exposed miners and estimation of risk from indoor exposure. J Natl Cancer Inst 1995; 87:817–827.
9. Mellander H, Enflo A. The alpha track method used in the Swedish radon epidemiology study. Radiat Prot Dosim 1992; 45:65–71.
10. Weinberg CR, Moledor ES, Umbach DM, Sandler DP. Imputation for exposure histories with gaps, under an excess relative risk model. Epidemiology 1996; 7:490–497.
11. Nyberg F, Agrenius V, Svartengren K, Svensson C, Pershagen G. Environmental tobacco smoke and lung cancer in nonsmokers: does time since exposure play a role? Epidemiology 1998; 9:301–308.
12. Boffetta P, Kogevinas M, Simonato L, Wilbourn J, Saracci R. Current perspectives on occupational cancer. Int J Occup Environ Health 1995; 1:315–325.
13. Ahrens W, Merletti F. A standard tool for the analysis of occupational lung cancer in epidemiologic studies. Int J Occup Environ Health 1998; 4:236–240.
14. Nordisk yrkesklassificering. Stockholm, Sweden: Arbetsmarknadsstyrelsen, 1983.
15. Nyberg F, Gustavsson P, Järup L, Bellander T, Berglind N, Jakobsson R, Pershagen G. Urban air pollution and lung cancer in Stockholm. Epidemiology 2000; 11:487–495.
16. Nyberg F, Agudo A, Boffetta P, Fortes C, Gonzalez CA, Pershagen G. A European validation study of smoking and environmental tobacco smoke exposure in non-smoking lung cancer cases and controls. Cancer Causes Control 1998; 9:173–182.
17. Gustavsson P, Jakobsson R, Nyberg F, Pershagen G, Järup L, Schéele P. Occupational exposure and lung cancer risk: a population-based case-referent study in Sweden. Am J Epidemiol 2000; 152:32–40.
18. Pershagen G, Nyberg F. Luftföroreningar och lungcancer i Rönnskärsområdet. IMM-rapport 1/95. Stockholm: Institutet för miljömedicin, Karolinska Institutet, 1995.
19. Axelsson G, Liljeqvist T, Andersson L, Bergman B, Rylander R. Dietary factors and lung cancer among men in West Sweden. Int J Epidemiol 1996; 25:32–39.
20. Pershagen G, Åkerblom G, Axelson O, Clavensjö B, Damber L, Desai G, Enflo A, Lagarde F, Mellander H, Svartengren M, Swedjemark GA. Residential radon exposure and lung cancer in Sweden. N Engl J Med 1994; 330:159–164.
21. StataCorp. Stata Statistical Software. Release 6.0. College Station, TX: Stata Corp, 1999.
22. Preston DL, Lubin JH, Pierce DA. EPICURE risk regression and data analysis software. Seattle: HiroSoft International, 1990.
23. Breslow NE, Storer BE. General relative risk functions for case-control studies. Am J Epidemiol 1985; 122:149–162.
24. Breslow NE, Day NE. Statistical Methods in Cancer Research. The Analysis of Case-Control Studies. IARC Scientific Pub. No. 32. vol. 1. Lyon: International Agency for Research on Cancer, 1980.
25. Lubin JH. Models for the analysis of radon-exposed populations. Yale J Biol Med 1988; 61:195–214.
26. Thomas DC. General relative-risk models for survival time and matched case-control analysis. Biometrics 1981; 37:673–686.
27. Brookmeyer R, Lyang KY, Linet M. Matched case-control designs and overmatched analyses. Am J Epidemiol 1986; 124:693–701.
28. Neuhäuser M, Becher H. Improved odds ratio estimation by post hoc stratification of case-control data. Stat Med 1997; 16:993–1004.
29. Alavanja MCR, Lubin JH, Mahaffey JA, Brownson RC. Residential radon exposure and risk of lung cancer in Missouri. Am J Public Health 1999; 89:1042–1048.
30. Pershagen G, Liang ZH, Hrubec Z, Svensson C, Boice JD Jr. Residential radon exposure and lung cancer in Swedish women. Health Phys 1992; 63:179–186.
31. Blot WJ, Xu ZY, Boice JD Jr, Zhao DZ, Stone BJ, Sun J, Jing LB, Fraumeni JF Jr. Indoor radon and lung cancer in China. J Natl Cancer Inst 1990; 82:1025–1030.
32. Schoenberg JB, Klotz JB, Wilkox HB, Nicholls GP, Gil-del-Real MT, Stemhagen A, Mason TJ. Case-control study of residential radon and lung cancer among New Jersey women. Cancer Res 1990; 50:6520–6524.
33. Luebeck EG, Heidenreich WF, Hazelton WD, Paretzke HG, Moolgavkar SH. Biologically based analysis of the data for the Colorado uranium miners cohort: age, dose and dose-rate effects. Radiat Res 1999; 152:339–351.
34. Darby S, Whitley E, Silcocks P, Thakrar B, Green M, Lomas P, Miles J, Reeves G, Fearn T, Doll R. Risk of lung cancer associated with residential radon exposure in south-west England: a case-control study. Br J Cancer 1998; 78:394–408.
35. Auvinen A, Mäkeläinen I, Hakama M, Castrén O, Pukkala E, Reisbacka H, Rytömaa T. Indoor radon exposure and risk of lung cancer: a nested case-control study in Finland. J Natl Cancer Inst 1996; 88:966–972.
36. Alavanja MCR, Brownson RC, Lubin JH, Berger E, Chang J, Boice JD. Residential radon exposure and lung cancer among nonsmoking women. J Natl Cancer Inst 1994; 86:1829–1837.
37. Létourneau EG, Krewski D, Choi NW, Goddard MJ, McGregor RG, Zielinski JM, Du J. Case-control study of residential radon and lung cancer in Winnipeg, Manitoba, Canada. Am J Epidemiol 1994; 140:310–322.
38. Ruosteenoja E. Indoor radon and risk of lung cancer: an epidemiological study in Finland (Doctoral thesis). Helsinki: Finnish Center for Radiation and Nuclear Safety, 1991.
39. Lubin JH, Liang Z, Hrubec Z, Pershagen G, Schoenberg JB, Blot WJ, Klotz JB, Xu Z-Y, Boice JD Jr. Radon exposure in residences and lung cancer among women: combined analysis of three studies. Cancer Causes Control 1994; 5:114–128.
40. Bäverstam U, Lagarde F. The impact of measurement errors in exposure assessment for residential radon: a stochastic approach. In: Ron E, Hoffman FO, eds. Uncertainties in Radiation Dosimetry and Their Impact on Dose-Response Analysis. National Cancer Institute. NIH Pub. No.99–4541. Washington DC: U.S. Department of Health and Human Services, 1999; 132–138.
41. Lagarde F, Pershagen G, Åkerblom G, Axelson O, Bäverstam U, Damber L, Enflo A, Svartengren M, Swedjemark GA. Residential radon and lung cancer in Sweden: risk analysis accounting for imprecision in the exposure assessment. Health Phys 1997; 72:269–276.