Radon in homes - A possible cause of lung cancer

-A cause 25-34. An earlier case-referent study [Scand j work health 5 (1979) 10-151 has indicated a possible relationship between lung cancer and exposure to radon and radon daughters in dwellings. Indoor radon concentrations seem to depend on both building material and leakage of radon from the ground. This new study, in a rural area, is a further attempt to elucidate the etiology of lung cancer, taking into consideration type of house and ground conditions, as well as smoking habits. Although the choice of a rural study population helped to eliminate various confounding exposures in the urban environment, it limited the size of the study because of the rareness of lung cancer in rural populations. Long-term residents, 30 years or more in the same houses, were studied, and again an association was found between lung cancer and estimated exposure to radon and radon daughters in homes. The data also seem to indicate the possibility of a multiplicative effect between smoking and exposure to radon and radon daughters in homes, but there was also some confounding between these factors in the data.

Lung cancer has become one of the predominating forms of malignant disorders in men of the industrialized world, and it is rapidly approaching this position also for women (37). The relations between lung cancer, or more precisely bronchial carcinoma, and smoking habits are generally recognized today (43), and there are suggestions of an association with passive smoking as we11 (15,27,46). However, some of the observations on the role of smoking in the causation of lung cancer are not quite consistent (and indicate that also other important factors are involved), and there have also been some more or less wellknown challanges of the smoking theory from time to time (13,22,38).
A number of occupational factors has been associated with lung cancer ( 4 9 , but they d o not seem to account fully for the increased incidence in urban areas as compared to rural districts (47). Furthermore the urban-rural gradient of lung cancer incidence exists even after allowance has been made for smoking habits (24), ie, it has been found among nonsmokers and within the various smoking categories (16,17,18).
Quite recently interest has focused on another potential and widespread risk factor for lung cancer, namely, exposure t o indoor levels of radon and radon daughters, as such exposure has presumably increased over past decades as central heating by radiators has replaced furnaces, etc, and a further increase is likely to result from efforts to save energy through the tightening of houses and decreasing ventilation (12,19,33,42,44).
Radon is a noble gas, originating from the decay of uranium through radium; it also decays itself by emitting alpha radiation into isotopes of lead, bismuth, and polonium, which are the so-called shortlived radon daughters, two of which emit alpha radiation. The occurrence of indoor radon and radon daughters is due t o a leakage of radon from the ground, but building materials also contribute, especially if the foundation is on gravel or alum shale. Furthermore radon is given off from tap water if it originates from ground water. Since radon has proved to be a pulmonary carcinogen in studies of underground miners (3), it is of crucial interest to determine whether or not an effect of low levels in dwellings can be epidemiologically demonstrated.
In an earlier case-referent study on the possible impact of exposure t o radon and its daughters in dwellings (5), we found an increased mortality from lung cancer among people in stone houses, where concentrations of radon and radon daughters tend to be higher (than in wooden houses) (28). In a n American study measurements of radon concentrations in the homes of lung cancer cases were found to be higher than the levels in the homes of the referents (controls), the result being an odds ratio of slightly over 2 (Radford, personal communication). A correlation study from the state of Maine in the United States has shown a positive and significant correlation between lung cancer and presumably high radon levels in tap water, as related to high radon levels in well water (26). Another correlation study from the state of Iowa has also indicated a relationship between the levels of radium-226 in drinking water and incidence of cancer of the lung and bladder among males and of cancer of the lung and breast among females (1 1).
An additional study (41) on lung cancer rates and housing characteristics, ie, constructions and building materials, did not disclose significant findings of any such relations, but the rates were highest in the context of concrete houses and houses built on a concrete slab. In Sweden lung cancer incidences on a county level have been found to correlate with background gamma radiation, which in turn can be as- sumed to reflect the levels of radon daughters in dwellings reasonably well (21).
These various observations have a more or less preliminary character, which calls for additional research in this field, and the present study was undertaken on the island of Oeland in the Baltic Sea to elucidate further a possible relation between indoor radon daughter exposure and lung cancer. Some preliminary results have already been presented (6), but, thanks to additional economic support, it has now been possible to obtain questionnaire information about the houses and to undertake radon daughter measurements. The smoking habits of the subjects have also been procured and a further check was made of medical records.

Material and methods
The island of Oeland, 140 km in length and up to 15 km in breadth, was chosen for the study because it is a rural area with particular geologic features, providing a rather unique study situation. The ground is limestone with uranium-containing alum shale underneath. But the layers of limestone and shale are slightly tilted towards the east, and the alum shale appears in the ground in a narrow strip (only some hundred metres in breadth) on the west side of the southern two-thirds of the island from a few hundred metres to about 2 km from the seashore (figure 1). This alum shale zone gives rise to an elevated level of gamma radiation exceeding 30 fiR/h (77.4 x C/kg. h), from the ground, and a somewhat increased leakage of radon into the air could be expected to occur. It is also very likely therefore that the indoor radon daughter levels in the houses on this particular ground are increased.
The population of the southern part of the island is peculiarly distributed, namely, in two strips, ie, part living along (within and outside) the alum shale zone on the western side and part residing on the eastern side of the island, whereas the inland of the island is sparsely populated (being a steppe-like area and forest landthis specific pattern is not followed in the northern third of the island, however). The total population of the island was about 20,000 individuals during the study period, 1960-1978, and there are three small towns with some 2,000-3,000 inhabitants each, all on the west coast.
As in other rural populations lung cancer was a relatively rare disease among the inhabitants, who made their living from farming, with associated services and tourism, or were employed (only a few hundred people) in small factories (cement and lightweight concrete production, sugar-beet refining and limestone quarrying). The plan was to undertake a case-referent study, but for economic reasons the study had to be built up step-by-step, evaluating the reasons and possibilities to proceed further. Thus a first analysis of the data was undertaken which showed a higher mortality from lung cancer in the western and northern parishes than in the eastern area (table 1). Since this difference could be related to the somewhat more urbanized structure of the west coast, but also was in accordance with the hypothesis of an increased lung cancer risk due to a higher background level of radon, the original plan of undertaking a case-referent study seemed necessary to evaluate the situation more definitively. Thereby it seemed reasonable to restrict the study to only that sector of the population which had been stable and had lived in the same houses for a very long time.

Selection of subjects
The official registers of deaths and burials in the 34 parishes of the island were merged and utilized as one single source register of subjects for this case-referent study. In all, 6,288 persons had died on the island during the 19-year study period, and, of these deaths, 69 were registered as due to lung cancer, all of persons 40 or more years of age. This registration was checked through a comparison with the medical files, and as cases those subjects were finally chosen who had died from malignant tumors of the lung [International Classification of Diseases (ICD) 1621. With the preliminary aim of attaining a one-to-six casereferent ratio, a random sample of 414 individuals was initially drawn as referents from the source register; thereby only those persons with noncancer diagnoses on their death certificates were enrolled.
After the material was restricted to include only those cases and referents (men as well as women) who were above 39 years of age and who had lived in the same house for 30 years or more prior to death, and not including cases and referents who had lived in apartments, 23 cases and 202 referents were included in the analysis. Individuals living in apartments were not included because of the difficulties involved in assessing exposure categories in this context.

Assessment of exposure and smoking habits
The addresses of the cases and referents, as given in the source register of deaths and burials, were located on large-scale economic maps, which permitted the geographic identification of isolated houses. "Geo-Radiation maps," originally produced for uranium ore prospecting, were used to determine whether or not a house was situated on radiation ground.
The final assessment of exposure was then obtained in two ways, ie, ocular classification of the houses and measurements of radon daughters by alpha sensitive film. For the classification of the houses with regard to exposure, every house was visited and, from judgements of the outside, building material, type of house, and ground conditions were taken into account. This classification also involved photographing the houses, and it was undertaken without any consideration of whether or not the house had belonged to a case or a referent. Three categories were used: the reference (category 0) being wooden houses without a basement on normal ground; category 2 being wooden houses with a basement on radiation ground (the alum shale zone) or stone, brick and plaster houses with a basement on any ground (alternatively without a basement on radiation ground); and exposure category 1 comprising all other types of houses, ie, also wooden houses without a basement but on radiation ground.
Questionnaires were also administered blindly to the next-of-kin of the subjects or present house owners, and the collected data were used for supplementary information. For a few houses the original classification had to be slightly adjusted according to information obtained about the existence of basements. The questionnaires also referred to major renovations, major changes in ventilation, and the occupations and smoking habits of the subjects.
Measurements of radon daughter concentrations were undertaken together with the local public-health authorities, and were made in as many houses as possible, owners permitting (86 %). Kodak LR 115 type I1 cellulose nitrate film was placed about 1.5-2 m above the livingroom floor. Three films per house were exposed during one month in March-April and during three months around the period April-June. The films were analyzed by the Department of Radiophysics at the University of Lund, according to principles described elsewhere (40). These analyses were performed without any knowledge of whether the house had belonged to a case or a referent. Because the three months of measurements went into the summer season, a period with much better ventilation, the one-month measurements were regarded as reflecting the average exposure situation better and were therefore used in the analysis. (This was a suggestion made by the Department of Radiophysics before any evaluation had started.) The categorization of the exposure levels was made with the fact in view that the average radon daughter concentration in Swedish houses seems to be about 50 Bq/m3, but also with regard to the generally held belief that film measurements are uncertain at levels below 100 Bq/m3. Furthermore there was a need to obtain a reasonable number of individuals at the highest exposure level to make the analysis meaningful. Thus the scale was set up as < 50 Bq/m3, 50-149 Bq/m3 and 1 150 Bq/m3. Information about, smoking habits was obtained for the cases and referents through the aforementioned questionnaires. For a few subjects there was a supplementary telephone interview made by a local dean. People who had stopped smoking 15 years or more prior to death were considered nonsmokers.

Statistical methods
The calculation of chi-square values and the estimatioh of the overall odds ratios, equaling the incidence density ratios with the design applied (4), were based on the Mantel-Haenszel procedures (32). The requirement of minimum expected cell size, ie, that there be a potential variation of at least 5 on each side of the expected total across the tables for a particular cell entry, was determined with a method described by Mantel & Fleiss (31). Trends were tested by the Mantel extension of the Mantel-Haenszel test, scoring the exposure categories as 0, 1, and 2 (30). The principles applied for determining the standardized rate ratios have been outlined by Miettinen (34,35), together with the method used for calculating the approximate (test-based) confidence interval of the rate ratios (36). Utilizing the GLIM (general linear interactive modeling) package (9), a logistic regression analysis was also applied, particularly for elucidating potential confounding and effect modification.

Results
As a result of the classification procedure (table 2), 10 of the 23 cases, or 43 Yo, and 125 of the 202 refer-ents, or 62 %, were found to belong to exposure category 0 (reference category). Since seven of all the cases (30 To) had lived in houses in exposure category 2 versus 22 of the referents (11 Vo When smoking habits were allowed for, the material was reduced by one case (4 %) and 24 referents (12 %) due to missing information. Stratification on age, gender, and smoking (table 3) reduced the Mantel-Haenszel point estimates somewhat, ie, for exposure category 1 + 2 the rate ratio amounted to 1.8 with a 90 VO confidence interval of 0.9-3.9 and for exposure category 2 it was 3.5 with a 90 VO confidence interval of 1.3-9.2. The multivariate logistic analysis resulted in almost identical figures, ie, a rate ratio of 3.9 and a 90 070 confidence interval of 1.5-10.0.
There was a slight positive confounding effect from smoking, 15 % of the referents who smoked belonging to exposure category 0 as compared to 27 % in exposure category 1 and 26 % in exposure category 2. When smoking was considered as the determinant of lung cancer and stratification was made on the basis of the various house categories, the risk ratio for smokers versus nonsmokers was 2.8 with a 90 070 confidence interval of 0.9-5.5. From table 3 it was also possible to obtain some information about potential interaction between smoking and radon daughter exposure. A comparison of smokers in exposure categories 1 and 2 to nonsmokers in exposure category 0 revealed a rate ratio of 3.6 and 9.3, respectively, a result indicating a more than additive effect from these two factors. The same conclusion could be drawn from the multivariate logistic analysis showing a multiplicative effect between radon daughter exposure and smoking.
Measurements of radon daughter levels were made in all but 3 case (9 Yo) and 29 referent (14 To) houses. As shown in table 4, the measurements were in reasonably good agreement with the applied exposure classification. In the different exposure categories, 86 YO of the houses in category 0 were measured, 84 070 in category 1, and 90 Vo in category 2.
The results obtained from the alpha sensitive films have been used in assessing the exposure levels for those cases and referents with known smoking habits (table 5). Eight cases (42 Yo) and 102 referents (64 To) belonged to the lowest exposure level (reference level) and three cases (16 Vo) had lived in houses with the highest exposure level (r 150 Bq/m3) versus seven referents (4 Vo). Consequently, the crude rate ratio they strengthen the idea of a possible relationship was 4.9 and the Mantel-Haenszel point estimate 5.1 between lung cancer and exposure to radon and with a 90 % confidence interval of 1.4-18.5. The radon daughters in dwellings. With regard to the corresponding figure for all the exposure levels above number of cases that were obtainable, the size of the 50 Bq/m3 was 2.8 for the crude rate ratio, and the study is small, and the Mantel-Haenszel chi-square Mantel-Haenszel point estimate was 2.7 with a 90 % test was found to be somewhat suboptimal according confidence interval of 1.1-6.4. The multivariate to the evaluation suggested by Mantel & Fleiss (31), logistic analysis ended up in a rate ratio of 4.7 and a ie, when the single exposure categories were consid-90 % confidence interval. of 1.2-18.5. The Mantel ered, but was acceptable for dichotomized data (ie, extension for trend resulted in a chi-square value of exposure categories 1 and 2 taken together). The 4.73 (p 1 0.02; one-tailed).
results are also somewhat suggestive that a lung cancer hazard could exist already at exposure levels Discussion around or below 150 Bq/m3, although the study is by no means conclusive in this respect. The results of this investigation are fairly well in The acquisition of cases and referents from the accordance with those of an earlier study (5), and parish registers could perhaps be suspected of not Mantel-extension X,2 = 5.34 a 0 = wood houses without a basement on normal ground; 1 = all types of houses not included in category 0 or category 2; 2 = wooden houses with a basement on radiation ground or stone, brick and plaster houses with a basement on any type of ground (alternatively without a basement on radiation ground).
being entirely satisfactory owing to the incompletecates were compared with those in the medical files, ness of such registers. However, the registers of there were five cases which could not finally be acdeaths and burials in Sweden usually contain full cepted as primary lung cancer and therefore the numinformation about the deaths as based on fairly ber of cases is fewer in this presentation than in the complete diagnoses from the death certificates. preliminary report of the same material (6). Nineteen However, when the diagnoses of the death certifiadditional referents have also been included that were

3.5
1.8 90 % confidence interval 0.5-3.0 1.3-9.2 0.9-3.9 Mantel-extension X,2 = 3.88 a 0 = wood houses without a basement on normal ground, 1 = all types of houses not included in category 0 or category 2, not in the earlier presentation due to the new entries for identification that were obtained through the questionnaire information.
In the selection of referents, only those who had died of nonmalignant diseases were considered, since background radiation not only might cause the disease under study (lung cancer) but also other cancers (I), and the inclusion of cancer diagnoses among the referents could have given a false impression of the exposure frequency in the various categories of the study base. This might perhaps be the case particularly for persons of younger ages, judging from experiences indicating a relation between acute myeloid leukemia and background radiation (23), especially since a correlation of radon daughter exposure and exposure to gamma radiation can be expected. It needs be pointed out perhaps that the applied exclusion of cancer diagnoses largely explains the change of the case-referent ratio from the original one-to-six to one of almost one-to-nine, since people with cancer tend to die at relatively younger ages, the possibility to fulfill the inclusion criteria of having lived in the same house for 30 years or more prior to death thereby having been reduced.
The validity of the information about smoking habits, as obtained through interviews with next-ofkin, may be criticized with regard to a "memory bias." According to recent experiences, however, the assessment of exposure through interviews seems to be acceptable when interviews are compared to other sources as far as occupational exposures and smoking histories are concerned (25,39). It is also known (14) that the number of smokers is generally low in rural areas, and in this study only 20 Yo of the referents were smokers. The exclusion of all cancers, some of them being related to smoking, would tend to give a lower frequency of smokers among the referents; such a procedure would have no impact on the main results of the study as long as this underfrequency is symmetrically related to the exposure categories however. This phenomenon might also be balanced by an overrepresentation of smokers with other specific diagnoses such as cardiovascular disease. Among the male cases, 60 % were smokers, but none of the female cases smoked. In the assessment of smoking habits, there was a dropout of 11 Vo of the total study population. Regarding only the referents, the relative dropouts were 12 % in exposure category 0, 11 Yo in category 1 and 14 Yo in category 2. In a reanalysis of the data under the assumption that all the referents with missing smoking information were nonsmokers whereas the missing case was a smoker, the crude rate ratio, as well as the Mantel-Haenszel point estimate, was certainly somewhat reduced, but there was still a numerically increased rate ratio for exposure category 2.
The alpha sensitive Kodak film (as well as other similar methods of measurement) has been criticized for insensitiveness and dispersion, but nevertheless it seems reasonable to believe that the method should be able to differentiate well enough between houses with low levels, formally taken as < 50 Bq/m3 in this study, and houses with higher levels, or formally r 150 Bq/m3. Lung cancer has been associated with exposure to pesticides (10) and, since farming is mainly of grain and vegetables in the southwest of the island, ie, along the alum shale strip, and the raising of livestock is more predominant in other areas, the possibility of confounding in this respect was also considered. To allow for this potential source of confounding but also for possible differences in smoking habits related to the more urbanized structure of the west coast or other "unknown" confounding factors as an explanation for the difference in lung cancer mortality between the eastern and western part of the island, an analysis was made also after restriction to the southwest only. The same relationship between lung cancer and the exposure categories was found, although for a limited number of individuals. Regarding occupations, the information from the questionnaires gave no indication of any association with lung cancer. By stratification according to wooden and nonwooden houses, the lung cancer risk was found to be higher on radiation ground (the crude rate ratio being 3.7); this finding is not surprising in view of the anticipated contribution of radon from the ground.
Finally the interaction between smoking and radon daughter exposure should be somewhat further discussed. Thus the findings in this study are in accordance with some of the studies of mining populations, particularly from uranium mines (2), where also a Table 4. Radon daughter concentrations with regard to radiation ground and exposure categorization; three films per house were exposed during one month.  more or less multiplicative effect has appeared. However, there is some inconsistancy with a coupre of mining studies (7, 20), in which a merely additive effect of radon daughter exposure and smoking was found. The reason could be that ordinary people are not exposed to the same amount of dust and other irritants as miners, particularly in older mines, and that these irritants together with smoking tend to increase the thickness of the mucous sheath and there-by somewhat protect against the very short-ranging alpha radiation. [For a further discussion see the article by Axelson & Sundell (8).] Thus there is not necessarily any direct contradiction present between our results and the mining studies, especially as the induction-latency periods seem to be shorter for the smoking miners than for their nonsmoking colleagues (2, 7), a phenomenon which seems consistent with a promotive effect for smoking.

Radon daughter concentrations Bqlm3 EERa
In conclusion, the findings presented in our present study, along with the results of the earlier study (5) and the general knowledge about the relationship between radon daughter exposure and lung cancer, seem to suggest that exposure to radon and radon daughters in dwellings may play a n important etiologic role for lung cancer in the general population. Therefore, the risk estimates obtained for smoking might be influenced or modified, or even somewhat confounded by indoor radon daughter exposure, and this possibility might help explain some of the irregularities reported for the relationship between smoking and lung cancer