Residential radon exposure, diet and lung cancer: A case-control study in a Mediterranean region
Article first published online: 11 JAN 2005
Copyright © 2004 Wiley-Liss, Inc.
International Journal of Cancer
Volume 114, Issue 6, pages 983–991, 10 May 2005
How to Cite
Bochicchio, F., Forastiere, F., Farchi, S., Quarto, M. and Axelson, O. (2005), Residential radon exposure, diet and lung cancer: A case-control study in a Mediterranean region. Int. J. Cancer, 114: 983–991. doi: 10.1002/ijc.20799
- Issue published online: 11 MAR 2005
- Article first published online: 11 JAN 2005
- Manuscript Accepted: 10 AUG 2004
- Manuscript Received: 16 MAR 2004
- European Union Nuclear Fission Safety Program. Grant Number: F14P-CT96-0055
- case-control study;
- lung cancer;
We performed a case-control study in Lazio, a region in central Italy characterized by high levels of indoor radon, Mediterranean climate and diet. Cases (384) and controls (404) aged 35–90 years were recruited in the hospital. Detailed information regarding smoking, diet and other risk factors were collected by direct interview. Residential history during the 30-year period ending 5 years before enrolment was ascertained. In each dwelling, radon detectors were placed in both the main bedroom and the living room for 2 consecutive 6-month periods. We computed odds ratios (ORs) and 95% confidence intervals (CIs) for time-weighted radon concentrations using both categorical and continuous unconditional logistic regression analysis and adjusting for smoking, diet and other variables. Radon measurements were available from 89% and 91% of the time period for cases and controls, respectively. The adjusted ORs were 1.30 (1.03–1.64), 1.48 (1.08–2.02), 1.49 (0.82–2.71) and 2.89 (0.45–18.6) for 50–99, 100–199, 200–399 and 400+ Bq/m3, respectively, compared with 0−49 Bq/m3 (OR = 1; 0.56–1.79). The excess odds ratio (EOR) per 100 Bq/m3 was 0.14 (−0.11, 0.46) for all subjects, 0.24 (−0.09, 0.70) for subjects with complete radon measurements and 0.30 (−0.08, 0.82) for subjects who had lived in 1 or 2 dwellings. There was a tendency of higher risk estimates among subjects with low-medium consumption of dietary antioxidants (EOR = 0.32; −0.19, 1.16) and for adenocarcinoma, small cell and epidermoid cancers. This study indicates an association, although generally not statistically significant, between residential radon and lung cancer with both categorical and continuous analyses. Subjects with presumably lower uncertainty in the exposure assessment showed a higher risk. Dietary antioxidants may act as an effect modifier. © 2004 Wiley-Liss, Inc.
Cohorts studies among miners provide ample evidence that exposure to radon and its decay products increases the risk of lung cancer.1, 2 Therefore, interest has been focused on indoor radon and, based on extrapolations from miner studies, it has been estimated that indoor radon poses a substantial lung cancer risk, i.e., about 5–20% of the total lung cancer burden.2, 3, 4 However, uncertainties in this extrapolation2, 5 have motivated epidemiologic studies in order to evaluate directly the potential lung cancer risk to the general population. In early small-sized studies, estimates of exposure to indoor radon were based on proxies such as the type of houses; geologic features were also considered, and some measurements of radon were finally introduced.6, 7, 8 In later studies, broader population groups were included with more extensive direct measurements of current radon concentration,9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28 along with measures of surface concentrations of 210Po implanted on glass surfaces to estimate past radon concentration.18, 29
The results of these epidemiologic studies on the lung cancer risk associated with environmental exposure to radon have been reviewed elsewhere.2, 30, 31, 32, 33, 34, 35, 36 A general methodologic problem affecting all of these studies is the effect of radon exposure uncertainties on the estimated risks. When such uncertainties are accounted for in the analysis, the estimated relative risk can increase by about 50% or more,17, 37, 38 but the correction necessarily widens the confidence interval.39, 40
None of the European studies has been completely carried out in a Mediterranean area. The mild climate of this area can affect indoor ventilation, the ratio between radon and its progeny concentrations and the aerosol distribution size. All of these factors have an effect on the lung dose due to inhalation of radon and its progeny and therefore on the risk.2, 41 Moreover, the Mediterranean diet is rich in antioxidants from fruit and vegetables and has been proven to be a protective factor for lung cancer.42, 43 The present study aimed at contributing to the European data set with an additional study from Lazio, Italy, a region with Mediterranean climate and diet. Lazio, which has about 5 million inhabitants, including Rome (with about 3 million inhabitants), is characterized by a volcanic geology and by a large use of building materials of volcanic origin, i.e., local tuff, with high exhalation of radon.44 The residential radon concentration measured in this region, in the framework of the Italian national radon survey, was found to be relatively high (average = 119 Bq/m3; geometric mean = 93 Bq/m3; geometric standard deviation = 1.9).45, 46
The objective of this study was to evaluate the degree of association between exposure to indoor radon and lung cancer in Lazio while considering the potential role of several determinants of lung cancer, including diet.
Material and methods
Selection of cases and controls and interview
The present study was conducted as an extension of the International Agency for Research on Cancer (IARC) collaborative initiative to investigate the role of environmental tobacco smoke,47 active smoking48 and markers of DNA damage and genetic susceptibility on lung cancer.49 Additional details of this study have been reported in a publication on dietary factors and lung cancer.42
This case-control study recruited subjects from one of the main hospitals of Rome (Forlanini-S. Camillo Hospital). Eligible cases were Caucasian adults ranging from 35 to 90 years of age who had lived in the Lazio region (Rome included) for at least 25 years and admitted to the hospital between November 1993 and June 1996 for suspected lung cancer to be further evaluated through diagnostic bronchoscopy. Eligible cases were asked for their consent to be interviewed before the diagnostic bronchoscopy.
Controls were frequency-matched to the cases by sex (1:1 for males; 1:2 for females) and age (in 5-year strata) and selected among patients admitted to the same hospital during the study period from the following wards: general surgery, orthopedics, ear, nose and throat (ENT), as well as general medicine. Subjects admitted to the hospital because of conditions that are somehow related to either smoking or dietary habits, such as many cancers, respiratory diseases, diabetes, cardiovascular, digestive and renal diseases, were not included in the controls. A balance between different diagnoses was kept when sampling controls.
Two trained assistants interviewed the patients using structured questionnaires regarding demographic data, smoking habits, occupational exposure to carcinogens, environmental tobacco smoke and diet. A few weeks after enrollment and the interview of the suspected lung cancer cases, all the relevant findings from bronchoscopy, thoracic surgery, pathology and other medical records from the hospital were examined for confirmation of the suspected lung cancer diagnosis. Histologic or cytologic evidence of a primary lung cancer was searched. A nonsmoker was defined as any subject who had smoked less than 400 cigarettes in her or his lifetime. An ex-smoker was defined as a smoker who had given up the habit at least 1 year before the interview. Subjects were classified as exposed to known (list A) or suspected (list B) occupational lung carcinogens on the basis of their positive response to a checklist of specific exposures/occupations.50
The Italian version of the EPIC food-frequency questionnaire (a self-administered questionnaire designed and used in the ongoing European Study of Diet and Cancer51) was used to assess dietary habits.42 Food items were subdivided into related groups and subgroups based on the type of nutrient they contain. For each individual food or food group, 3 categories of low, medium and high consumption were defined.
Residential histories and radon concentration measurements
The radon study began in 1996 when nearly all the interviews had been completed. In order to select the individuals for the radon study, the residential history of each study subject during the 5–34 years preceding enrollment (the period of exposure to indoor radon that we assumed as relevant to the risk of lung cancer) was collected from the municipal registry of the last residence. For each dwelling occupied for 1 year or more, the complete address and the relevant dates were collected from the municipal registry and through a short questionnaire administered (by mail or telephone) to cases and controls (or to the next-of-kin). In order to resolve some inconsistencies, each individual residential history was confirmed by the subject (or the next-of-kin) during the home visit for the placement of the radon detectors.
For apartments, in the case of a refusal or inability to contact the present occupant of the residence, the radon concentration in a proxy dwelling in the same building was measured. A proxy dwelling was preferably on the same floor as the target dwelling; otherwise, the closest dwelling above or below was selected. In order to evaluate the effect of using proxy dwellings, we performed a parallel measurement of radon in a sample of about 60 pairs of target and proxy dwellings. There were generally only minor differences in the measured radon concentrations.
Particular attention has been devoted to radon concentration measurements. One passive device for measuring radon concentration (hereafter “radon device”) was placed in both the main bedroom and the living room of each dwelling occupied by the study subjects in the relevant 5–34-year period before enrollment. Moreover, the radon devices were exposed for 2 consecutive 6-month periods in order to reduce the random measurement error, the effect of detector saturation and losses. Each radon device (described in more detail elsewhere44, 52) contains 2 LR115 detectors in a closed-type configuration. This unique characteristic allowed an internal quality control analysis for each radon device; the results of this analysis as well as the details of the quality assurance program have been reported elsewhere.45, 52 For each radon device, the average result of the 2 contained LR115 detectors was used to assess the radon exposure. The first radon device was installed in September 1997 and the last one was retrieved in November 1999. Other details on the design of the radon part of this study are reported elsewhere.46
Assessment of time-weighted average radon concentration
For each subject, the time-weighted average (TWA) radon concentration was estimated using residences both with and without radon measurements. The weights were equal to the years spent in a particular dwelling. For each address measured, the annual radon concentration was obtained by averaging the data of the 2 6-month periods and of the 2 rooms, weighting for the actual exposure times of the 2 periods. For addresses monitored for only one 6-month period, the average annual concentration was estimated using a seasonal correction factor derived from the set of complete measurements. For dwellings with no radon measurements, an imputation strategy was applied based on Weinberg et al.,53 using the average radon concentrations measured among the controls living in the same area of Lazio (5 districts in Rome and the 5 counties of the Lazio region, whose averages ranged from 56 to 246 Bq/m3). For residences outside Lazio, the regional average values from the Italian national survey were used, ranging from 25 to 124 Bq/m3.45 For residences abroad, a radon concentration in the range of 11–70 Bq/m3 was assumed on the basis of relevant published data.4, 54
To evaluate the effect of indoor radon on lung cancer, unconditional logistic regression analysis was used to calculate odds ratios (ORs) and 95% confidence intervals (95% CIs). We evaluated the role of several potential predictors of lung cancer. Sex and age were considered in the design. An interaction term (sex × age) was introduced to accommodate the slightly different age distribution by sex among cases and controls. Stratification of age in 5 years instead of 10 did not significantly change the risk estimates but increased their variability so that 10-year strata were used. Because the referral pattern of cases and controls was different for residents in Rome and in the northern part of the region compared to residents of the rest of Lazio, adjustment was also made for area of residence.
Years of school attendance (as a variable with 3 categories) and exposure to occupations known to present an excess risk for lung cancer (exposed to list A and list B occupations vs. never exposed) were also considered. We considered 2 different approaches to control for smoking. In the first approach, 11 sex-specific categories were used: 1, never smokers; 2–5, ex-smokers for < 10 or ≥ 10 years who smoked < 25 or ≥ 25 cigarettes per day; 6–11, current smokers divided according to age at start (< 18 or ≥ 18 years) and number of cigarettes per day (< 15, 15–25, > 25). In the second approach, suggested by Leffondré et al.,55 we considered 3 smoking variables simultaneously in the logistic regression: an indicator for ever smoking (no, yes), cumulative cigarettes/day-years (continuous) and time since cessation of smoking (categorical). A variable cigarettes/day-years was calculated for both current and ex-smokers. Time since cessation was divided into 4 categories; current smokers and ex-smokers who had quit in the last 5 years were considered the reference category. A statistical interaction was suggested for ever smoking and we then included also a sex × ever-smoking interaction term. This second approach was used only in the sensitive analysis. Finally, the following 5 dietary variables, which in a previous analysis appeared to be related with a decreased lung cancer risk,42 were a priori considered as potential confounders: consumption of carrots, tomatoes, white meat, sage and exclusive use of olive oil.
The association of TWA radon exposure and lung cancer was first evaluated using the exposure variables in categories (0–49, 50–99, 100–199, 200–399, 400+ Bq/m3) and reporting the relative risk parameters using the floating absolute risks method proposed by Plummer.56 This method assigns a floated variance to each level of the risk factor and describes the uncertainty of risk without reference to the baseline level. To evaluate the effect of radon exposure as a continuous variable, the excess odds ratio (EOR) at 100 Bq/m3 compared to 0 Bq/m3 was estimated. The analysis was repeated considering 2 subsets of cases and controls with probable lower radon exposure uncertainty and larger exposure variability: subjects with all radon measurements available for the reference period, and subjects who lived in no more than 2 dwellings during the same period. Effect modification on the estimated EOR in the overall sample (and heterogeneity tests) was evaluated for sex, age, smoking and histologic type of lung cancer. Moreover, effect modification was also evaluated for consumption of carrots and tomatoes, food rich in antioxidants, and related to lung cancer in our as well as other studies. Finally, a sensitivity analysis was conducted to evaluate the robustness of the findings to different smoking adjustment and to other model specifications.
Table I shows the outcome of the selection process of the subjects under study. In all, 679 patients (74.0% of 918 originally identified as suspected lung cancer cases) and 443 controls (74.0% of the 599 control subjects approached) were interviewed. For reasons shown in Table I, we ended up with 384 cases and 404 controls in the analyses. In particular, 248 subjects had respiratory diseases instead of lung cancer as first suspected. Table I also shows the residential and the main demographic characteristics of cases and controls. Histologic or cytologic confirmation of the lung cancer diagnosis was available for 334 cases (87.0%) while there was no microscopic evidence for 50 cases (13.0%), and the diagnosis was based on clinical evidence only. The diseases more frequent among the controls were conditions of the sense organs (19.4%, hospitalized in the ENT ward), musculoskeletal diseases (17.7%), injury and poisoning hospitalized in the orthopedics ward (16.8%) and digestive diseases (11.5%).
|Subjects approached for the study||918||100.0||599||100.0|
|Not primary lung cancer||2481||27.0|
|Residential history not available||19||2.5||13||2.2|
|Lived in study area < 25 years||17||1.9||13||2.2|
|No radon measurement in the period of interest||9||1.0||10||1.7|
|Incomplete smoking data||2||0.2||3||0.5|
|Included in the analysis||384||100.0||404||100.0|
|Residence at interview|
|City of Rome, counties of Rieti and Viterbo||260||67.7||345||85.4|
|Counties of Rome, Latina and Frosinone||124||32.3||59||14.6|
Table II shows the coverage results of the radon measurement program. The mean number of residences recorded in the personal histories was similar between cases (2.8) and controls (2.9). Out of a total of 2,258 dwellings to be assessed (1,088 for cases and 1,170 for controls), radon concentration measurements were performed for 1,244 index dwellings (55.1%) and for 466 proxy dwellings (22.3%). The proxy dwellings were on the same floor in most of the cases, as required by the protocol. There were no major differences in the radon measurement coverage between cases and controls. In all, we measured 1,710 addresses (75.7% of the target) corresponding to an average of 2.1 residences for cases and 2.2 for controls. The dwellings not measured were generally inhabited for shorter periods (4.3 vs. 12.5 years), so that, within the 30-year reference period, measurements were available for an average of 27.0 years (26.8 for cases and 27.2 for controls), corresponding to an average coverage of 90% (89.3% for cases and 90.6% for controls). As expected, the most complete period was the most recent (5–14 years before enrolment), with a respective coverage of 95.9% and 96.3% for cases and controls. The total measurement duration was equal to the planned 6 + 6 month period for about 95% of the measured dwellings, while a measurement for only one 6-month period was obtained for the remaining 5% and the annual average was estimated as described above.
|Number of addresses reported||2.8||1–8||2.9||1–9|
|Number of addresses measured||2.1||1–7||2.2||1–6|
|Number of years with radon measurements||26.8||1–30||27.2||1–30|
|Percentage of years with radon measurements for different exposure windows|
|5–14 years before enrolment||95.9||10–100||96.3||10–100|
|15–24 years before enrolment||92.6||10–100||94.6||10–100|
|25–34 years before enrolment||79.5||10–100||81.0||10–100|
|5–34 years before enrolment||89.3||3–100||90.6||3–100|
The distribution of TWA radon concentration has a log-normal shape and Table III reports the summary statistics for all cases and controls (subgroup 1), subjects with complete radon measurements (subgroup 2) and subjects who lived in no more than 2 dwellings (subgroup 3). Both the level and the variability of the exposure measures were higher among cases and controls in subgroups 2 and 3.
|All subjects, n||384||404|
|Average (range)||113 (28–529)||102 (28–456)|
|GM (GSD)||98 (1.67)||89 (1.67)|
|Median (25–75%)||94 (69–131)||83 (63–116)|
|Subjects with complete measurements, n||206||204|
|Average (range)||125 (28–529)||108 (29–456)|
|GM (GSD)||105 (1.76)||92 (1.72)|
|Median (25–75%)||100 (71–146)||85 (64–122)|
|Subjects lived in ≤ 2 residences, n||189||186|
|Average (range)||122 (28–529)||105 (29–456)|
|GM (GSD)||102 (1.77)||89 (1.72)|
|Median (25–75%)||95 (70–139)||83 (61–123)|
Tables IV and V report the odds ratios of the association of smoking (Table IV) and other variables (Table V) with lung cancer. Compared to never smokers, the risk of lung cancer for male increases with the number of cigarettes smoked per day and decreases with time since cessation among ex-smokers, although the confidence intervals are quite wide. The corresponding odds ratios for female smokers were generally lower and with less clear trend, but confidence intervals were large due to the low number of female subjects. After adjusting for smoking, no association was seen for occupational exposure to substances/conditions on list A, and there was only a slight tendency for risk to diminish as level of education increased. Table V also shows the smoking-adjusted odds ratio for the dietary variables. Protective effects, as already reported for a slightly larger data set,42 were seen for all the variables considered.
|nca/nco||OR||95% CI||nca/nco||OR||95% CI|
|Stop ≥ 10 years|
|< 25 cigarettes/day||46/71||4.1||1.6–10.7||3/11||1.2||0.3–4.8|
|≥ 25 cigarettes/day||25/18||10.8||3.7–31.6||0/0|
|Stop < 10 years|
|< 25 cigarettes/day||35/21||11.1||3.9–31.4||5/4||5.8||1.1–30.9|
|≥ 25 cigarettes/day||35/9||28.2||8.9–89.3||0/1|
|Age start ≥ 18 years|
|< 15 cigarettes/day||11/12||6.6||1.9–22.6||6/11||1.6||0.5–5.1|
|> 25 cigarettes/day||22/6||29.8||8.2–108||4/3||12.3||1.9–80.0|
|Age start <18 years|
|< 15 cigarettes/day||17/20||7.4||2.4–22.5||5/5||6.9||1.4–33.6|
|> 25 cigarettes/day||54/28||17.5||6.5–47.5||2/2||3.3||0.4–26.1|
|< 9 years||251/227||1.00|
|> 13 years||50/75||0.77||0.48–1.22|
|Consumption of carrots|
|Consumption of tomatoes|
|Consumption of white meat|
|Exclusive use of olive oil|
Table VI reports the results of the analysis of the association between lung cancer and TWA radon concentration in categories. The first series of odds ratios is adjusted for sex, age, sex × age, residence and smoking variables, as occupational exposure and education did not exert a confounding effect. In the second series of odds ratios, dietary variables were also considered in the multiple adjustment. There was a smooth increase in the risk of lung cancer as TWA radon concentration increases, with higher values of the ORs (and the p-value for trend) when adjustment for diet was made. On that basis, all the additional results are presented with dietary adjustment. The strength of the association tended to be stronger when the analysis was restricted to those with complete radon measurements and among those with 1 or 2 residences. In the latter group, the measured radon exposure is expected to be more closely related to the actual exposure of the subject due to fewer changes in living habits and dwelling restructures.
|TWA radon concentration (Bq/m3)||nca/nco||OR1||95% CI||p (trend)||OR2||95% CI||p (trend)|
|Subjects with complete radon measurements|
|Subjects lived in ≤ 2 residences|
With continuous unconditional logistic regression analysis, the EOR per 100 Bq/m3 was 0.14 in the overall sample. Table VII reports the results of the subgroup analyses. Due to small sample size in some categories for females, control for smoking was done omitting sex stratification in order to reduce the loss of female subjects in the analysis. The EOR per 100 Bq/m3 increased to 0.24 and 0.30 in the 2 subgroups with expected better radon exposure assessment compared with 0.12 (−0.12, 0.42), when stratification of smoking by sex was omitted in the main analysis). Effect modification by sex, age and smoking was far from statistical significance on the basis of the reported chi-square statistics. A larger effect was suggested among those with low-medium consumption of carrots and tomatoes (EOR = 0.32). When the analysis was done for histologic type, increased EORs were found especially for adenocarcinoma (0.36), but also for small cell cancer (0.22) and epidermoid cancer (0.19), although with large confidence intervals.
|Subgroups||nca/nco||EOR||95% CI||Heterogeneity test|
|All subjects||384/404||0.14||(−0.11, 0.46)|
|Complete radon measurements||206/204||0.24||(−0.09, 0.70)|
|Lived in ≤ 2 residences||189/186||0.30||(−0.08, 0.82)|
|Female||59/109||−0.07||(−0.48, 0.67)||p = 0.712|
|65+||215/174||0.18||(−0.16, 0.64)||p = 0.784|
|Ever||355/290||0.16||(−0.12, 0.51)||p = 0.493|
|Consumption of carrots and tomatoes|
|High||163/171||0.02||(−0.29, 0.48)||p = 0.287|
|Small cell||43/404||0.22||(−0.21, 0.89)|
|Clinical diagnosis only||50/404||−0.07||(−0.45, 0.58)|
Since the adjustment for smoking is important in a study on lung cancer, we considered in a sensitivity analysis the adjustment (described above) based on cigarettes/day-years as continuos variables. The EORs at 100 Bq/m3 for all the subjects, those with complete radon measurements and those who lived in 1 or 2 dwellings were 0.10 (−0.14, 0.41), 0.23 (−0.11, 0.69) and 0.28 (−0.10, 0.83), respectively. For the 2 latter subgroups, the results would be 0.32 (−0.05, 0.82) and 0.37 (−0.04, 0.96), respectively, when adjusting with the 11 sex-specific smoking categories; however, these results are partly due to the loss of female subjects discussed above.
In the sensitivity analysis, the effect of removing the considered confounders on the overall EOR, 0.14 (−011, 0.46) was evaluated. As indicated, a change in the estimated risk was obtained by removing diet from the model (EOR = 0.08; −0.15, 0.38). Given the design of the study, area of residence was an important factor (EOR = 0.34; 0.05, 0.70). No change was found EOR (0.14; −0.11, 0.47) when 5-year strata instead of 10-year strata were considered. Finally, we considered whether the higher effect we found for the restricted group of those who lived in 1 or 2 residences was due to some residual confounding, since some differences for smoking and education were found between this group and the other subjects. However, the result, already adjusted for smoking, remained practically unchanged when we added an adjustment for education (EOR = 0.31; −0.07, 0.86).
We found an association between indoor radon and lung cancer in Lazio, as shown by both the categorical and the continuous analysis. The increase was generally not statistically significant, although significance was reached for some radon exposure categories.
There are several strengths in the present study that should be noted. One, we have data on a large set of potential risk factors for lung cancer that were collected from living incident subjects and alive controls. Two, we ensured the accuracy of residential histories in order to avoid possible misclassifications. Three, the population studied is more stable in comparison to subjects investigated in other countries, especially in North America. Four, radon concentration was measured for 2 consecutive 6-month periods using multiple radon detectors (with a total of 8 detectors for each measured dwelling and an average of about 17 detectors per subject) to reduce measurement errors. The results of a quality control analysis on our radon measurements were very good compared to those of the only other published study.52, 57 Five, we had high measurement coverage of the 30-year period of exposure, thus reducing the uncertainties due to imputing missing values.
There are limitations of the study as well. One, the sample size is not very large, although comparable to some earlier investigations. Therefore, we did not have enough power to investigate some variables as effect modifiers. Two, the use of a single hospital, although it is one of the biggest in the region, to recruit subjects cannot guarantee that some bias was not introduced. Three, we had to rely on proxy dwellings (an apartment close to the index apartment) in a considerable percentage of the cases, thus introducing a potential error in the exposure estimate. However, our parallel survey in the index and in the proxy houses does not indicate that a large difference exists. Therefore, the use of proxy dwellings to estimate the missing value of radon concentration in many index dwellings has probably introduced a smaller error than imputation would have.
Our results indicate that the radon risk in this Mediterranean area is comparable to the risk estimates derived from previous studies carried out in countries with different climates, heating systems and living habits, most of which are typical of a colder climate as in continental Europe or in the Nordic countries. The EORs at 100 Bq/m3 estimated in previous studies (13 from Europe, 5 from North America and 2 from China) range from −0.06 to 1.48 (Table VIII). Most of these studies (17 out of 20) estimate an EOR greater than 0, although it was significant in only few studies. Our results for all subjects (EOR = 0.14; −0.11, 0.46) are similar to the results in some meta-analyses and pooled analyses. In particular, a meta-analysis of the first 8 studies gave a result of 0.12 (0.0, 0.3) for EOR.32 In a more recent meta-analysis, which included 6 more studies for a total of 14, an EOR of 0.06 (0.01, 0.10) was estimated.34 The EORs obtained in the recent pooled analyses of 2 Chinese studies and of 7 North American ones are quite similar: 0.13 (0.01, 0.36) and 0.11 (0.00, 0.28), respectively.58, 59 These similar results could imply that the dose per unit of radon exposure is not quite different between the studies, i.e., that some parameters affecting lung dose41, 60 could vary among studies but their combination does not change considerably, although the large confidence intervals could mask differences.
|Study area (year of publication)||nca||nco||Smoking status||CRn||EOR||(95% CI) at 100 Bq/m3||Comment|
|New Jersey, USA (1990)9||433||402||All||26||0.28||(−0.28, 0.97)|
|Shenyang, China (1990)10||308||356||All||118||−0.04||(−0.23, 0.19)|
|Stockholm, Sweden (1992)11||201||378||All||128||0.52||(−0.05, 1.54)|
|Sweden-I (1994)12||1,281||2,576||All||107||0.10||(0.01, 0.22)|
|Sweden-I (1997)37||1,281||2,576||All||0.17||(0.03, 0.37)||Adjusted for measurement error|
|Missouri-I, USA (1994)13||538||1,183||NS, ES||67||0.05||(−0.13, 0.24)||All subjects|
|197||1,183||NS, ES||0.47||(0.03, 1.40)||Living subjects only|
|Winnipeg, Canada (1994)14||738||738||All||120||−0.06||(−0.14, 0.05)|
|South Finland (1996)16||164||331||All||218||0.57||(0.27, 0.99)|
|Finland (1996, 1998)15||517||517||All||96||0.11||(−0.06, 0.31)|
|Southwestern England (1998)17||982||3,185||All||56||0.08||(−0.03, 0.20)||All subjects|
|982||3,185||All||52||0.12||(−0.05, 0.33)||Adjusted for measurement error|
|484||1,637||All||55||0.14||(0.01, 0.29)||Complete radon coverage|
|484||1,637||All||48||0.24||(−0.01, 0.56)||Adjusted for measurement error|
|Missouri-II, USA (1999)18||247||299||All||59||0.04||(−0.13, 0.57)||Contemporary radon measurements|
|372||471||All||65||0.63||(0.07, 1.93)||Retrospective radon measurements|
|Iowa, USA (2000)19||413||614||All||89||0.16||(−0.03, 0.61)||All subjects|
|283||614||All||0.33||(0.02, 1.23)||Living subjects only|
|Western Germany (2001)20||1,449||2,297||All||50||−0.02||(−0.18, 0.17)||All subjects, current home|
|365||595||All||60||0.13||(−0.12, 0.46)||Radon-prone areas|
|Sweden-II (2001)21||258||487||NS||79||0.28||(−0.05, 1.05)||All subjects|
|Sweden-II (2002)29||109||229||NS||90||0.33||(−0.12, 2.00)||Contemporary radon measurements|
|109||229||NS||83||0.75||(−0.04, 4.30)||Retrospective radon measurements|
|Trentino, Italy (2001)22||138||210||All||130||0.40||(−0.70, 5.60)|
|Pluton, Czech Rep. (2001)23||210||12,004||All||509||0.09||(0.02, 0.21)||Cohort study|
|Imst, Austria (2002)24||194||198||All||200||0.25||(0.08, 0.43)|
|Northwestern Spain (2002)25||163||241||All||130||1.48||(0.12, 4.48)|
|Gansu, China (2002)26, 38||768||1,659||All||228||0.16||(0.03, 0.40)||All subjects|
|768||1,659||All||0.29||(0.03, 1.04)||Adjusted for measurement error|
|463||1,143||All||0.23||(0.06, 0.57)||> 70% radon coverage only|
|463||1,143||All||0.65||(0.16, 3.04)||Adjusted for measurement error|
|Eastern Germany (2003)27||1,192||1,640||All||74||0.08||(−0.03, 0.20)||All subjects|
|427||536||All||0.09||(−0.06, 0.27)||Complete radon coverage|
|France (2004)28||552||1,103||All||148||0.04||(−0.01, 0.11)|
|Present study, Italy (2004)||384||404||All||107||0.14||(−0.11, 0.46)|
Analyses of restricted groups with presumably better radon exposure assessment provide indications of a stronger association in our study as well as in others. In our study, the analysis restricted to subjects with complete radon measurements and to subjects who inhabited 1 or 2 residences in the period of interest showed increased risk estimates—EOR = 0.24 and 0.30, respectively (Table VII)—probably due to both lower radon exposure uncertainty and higher geometric standard deviation of subject exposures (Table III). In the southwestern England study, the risk in the subgroup of subjects with complete radon measurements is higher and becomes statistically significant (Table VIII).17 A similar but stronger effect was observed in the Gansu study when subjects with at least 70% coverage of radon measurements are selected (Table VIII).38 In the pooled analysis of 2 Chinese studies, the risk tends to increase as the coverage of radon measurements increases, with greatest increase for subjects with complete coverage of 5–30-year exposure time window (EOR = 0.33; 0.07, 0.91).58 A similar tendency was observed in the pooled analysis of 7 North American studies: 0.21 (0.03, 0.50) for subjects for whom measured radon concentrations were available for at least 20 years within the period of interest (5–30 years before enrollment), and who had occupied at most 2 residences.59 The 2 published studies in which a retrospective technique was also used to evaluate past radon exposures showed higher risk estimates compared to the corresponding values obtained with contemporary radon dosimetry, suggesting lower misclassification in the retrospective exposure evaluation.18, 29 In the Missouri-I and Iowa studies, the risk estimated from the subgroup of living subjects was higher than for all the subjects and becomes statistically significant (Table VIII).13, 19 Uncertainty in ascertaining radon exposure generally tends to reduce significantly the observed risk. Our present estimates of EOR are not corrected for the effect of uncertainty in the radon exposures. In some studies (Sweden, southwestern England, western Germany, Gansu), the sources and the strength of such uncertainty has been analyzed and used to correct or to explain the observed results (Table VIII).17, 37, 38, 61 All these results confirm the importance of reducing radon exposure uncertainties due to their impact on risk estimates.
Our study had low power to detect effect modification and for subgroup analyses. Therefore, it is no surprise that all the heterogeneity tests we conducted were far from being statistically significant. However, there was an indication of a higher risk for some types of lung cancer, in particular adenocarcinoma and, to a less extent, small cell cancer and epidermoid cancer, than for other histologic types or clinical diagnosis only. Whereas a higher risk for small cell lung cancer was observed in other studies, no similar behavior was previously found for adenocarcinoma and it could be due to the low number of cases.2, 62 Particularly interesting is the stronger effect, although not statistically significant, observed in subjects with a low consumption of food rich in antioxidants (carotenoids and vitamin C in carrots and tomatoes). Both experimental and modeling data suggested a possible role of antioxidants to reduce part of the damage produced by α particles emitted by radon decay products.63, 64, 65 With the exception of the case-control study in never smokers in Germany,66 where no effect of indoor radon was found, no study to our knowledge has examined radon and diet in combination. Given that the role of diet in the etiology of lung cancer has been recognized, the issue of interaction between the 2 relevant determinants should be further investigated.
In summary, most of the epidemiologic studies available so far, including the present study, indicate a role of indoor radon exposure for lung cancer. This study, which is based on a very extensive radon concentration measurement program, also supports this role with both categorical and continuous analysis results. Moreover, it suggests that diet could be an effect modifier. The results of this study may play a significant role in the development of radon legislation in Mediterranean countries, where such regulations are generally lacking.67
Professor Olav Axelson, a leading scientist in the field of environmental epidemiology, passed away on 1 March 2004. From his experience derived from miner's studies, he was the first to postulate that residential radon exposure could pose a lung cancer risk.6 He was a brilliant scientist and a man of great integrity and the authors are grateful to have had the privilege of working closely with him. The authors also thank all those who have collaborated: G. Schmid and C. Perucci for their contribution to the study; C. Fortes for the design of the study regarding diet and corresponding data collection; D. Abeni and E. Rapiti for their contribution to implement the early phase of the radon study; F. Anatra and T. Trequattrini for patient recruitment; O. Catena and L. Scanavini for the interviews in the hospital; V. Goffredo and G. Piras for data entry; F. Sera for his contribution to data analysis; G. Grisanti and C. Nuccetelli for their contribution to the setup of the experimental apparatus and useful discussion on experimental aspects; A. Minella for the organization of radon device measurements; G. Moroni, M. Ampollini, M. Ciotti, M. Ceccarelli and F. Artizzu for manufacturing, etching, and track counting of radon detectors; A. Grisanti and F. Felici for their contribution to experimental measurements; S. Mallone, M. Piras, P. Compagnucci, S. Agostinoni, R. Fabriani, V. Fabriani, R. Giustini, P. Lorusso, B. Malchiodi, N. Mignolli, C. Monteleone, B. Plebani, P. Rega, M. Teodori, A. Trivini and G. Usai for the organization and the field work. A special thanks is due to all the study subjects and the inhabitants of all the selected dwellings.
- 1IARC. Monographs on the evaluation of carcinogenic risks to humans. Radon and man-made mineral fibres. vol. 43. Lyon: IARC, 1988.
- 2National Research Council. Health effects of exposure to radon. committee on health risks of exposure to radon: BEIR VI. Washington, DC: National Academy Press, 1999.
- 4WHO. Air quality guidelines for Europe, 2nd ed. Geneva: WHO, 2001.
- 23Study of lung cancer and residential radon in the Czech Republic. Cent Eur J Publ Health 2001; 9: 150–3., , , , .
- 24Radon und Lungenkrebs im Bezirk Imst/Österreich: Fortschritte in der Umweltmedizin. Landsberg am Lech: Ecomed Verlagsgesellschaft, 2002., , , , , , , , .
- 33IARC. Monographs on the evaluation of carcinogenic risks to humans. Some internally deposited radionuclides. vol. 78. Lyon, IARC, 2001.
- 38Adjusting lung cancer risks for temporal and spatial variations in radon concentration in dwellings in Gansu province, China. Radiat Res, in press., , , , , , , , , , , .
- 54United Nations Scientific Committee on the Effects of Atomic Radiation. Sources and effects of ionizing radiation, annex B: exposures from natural radiation sources. In: UNSCEAR 2000 report to the General Assembly, with scientific annexes: I, sources. New York: United Nations, 2000. 84–156.
- 59Risk of lung cancer in North America associated with residential radon. Epidemiology, in press., , , , , , , , , , , , et al.
- 67Radon legislation and national guidelines. SSI Rep 1999; 99: 18..