+7 (343) 287-57-36

620075, г. Екатеринбург
ул. К. Либкнехта, 8-Б

RUS ENG

test

Publications

Главная » Publications

RESIDENTIAL RADON AND LUNG CANCER: ANALYSIS OF DATA FROM THE THREE RUSSIAN MULTIFACTORIAL STUDIES

E.V.Polzik, V.L.Lezhnin, V.S.Kazantsev, I.V.Yarmoshenko

Institute of Industrial Ecology, UB RAS

The main unfavourable human health effect of radon exposure is lung cancer. Evidences of the carcinogenic effect were obtained in studies undertaken among miners of uranium and non-uranium mines [7,9,10,13,15]. At the same time, the evaluations of carcinogenic risk due to residential radon exposure are very controversial. Thus, according to the review of 13 European studies, completed by Darby et al. [5], not a single one of them has revealed a statistically significant positive relation between observed residential radon concentration and lung cancer development. Similar results are presented in other publications [4,11,14].

The basic reason for the remaining uncertainty, in our opinion, is application of the mono-factorial analysis methods by most of the researchers, which do not completely meet the complex requirements of the problem. In fact, cancer belongs to the category of diseases arising under the influence of many risk factors and, as a result, in order to single out the impact of one of them it is necessary either to eliminate the impact of many other factors or to use the methods of system analysis. Estimation of the results of published radon and lung cancer epidemiologic studies shows that their authors rather match the case and control groups by 2 main lung cancer risk factors – smoking status and sex. As a result, the insufficient elimination of confounding risk factors causes the above mentioned significant variability of results.

The meta-analysis of the published data or generalized analysis based on summarizing the data of many studies (pooling study) are considered to be main approaches to decrease the uncertainty of radiation risk estimates. The meta-analysis approach applied in the work of Lubin, Boice [12] summarized the results of 8 studies including 4263 persons with lung cancer and 6612 controls and the authors estimated the relative risk at radon concentration of 150 Bq/m³ as 1.14 with 95% confidence interval 1.0-1.3 [12]. Pavia et al [14] in their meta-analysis of 17 epidemiologic studies singled out higher relative risk – 1,24 (95% confidence interval 1.11-1.38) for the same value of 150 Bq/m³ radon concentration. In the largest published analysis summarising the initial results from 13 studies and including data for 7148 lung cancer patients and 14208 controls the relative risk of lung cancer increases by 0.08 per 100 Bq/m³ of radon concentration (95% confidence interval 0.03-0.16) [5]. As one can see from the above listed materials, the radiation risk estimates are still quite variable, so it is quite reasonable to suppose that mono-factorial studies, even based on thousands of cases, don’t allow to improve the reliability and statistical significance at indoor radon concentrations bellow 200 Bq/m3.

From our point of view, the correct approach to solving such an epidemiological problem is the use of multi-factorial analysis, which does not require the complex elimination of confounding risk factors. Application of this analysis method in the studies undertaken in the Russian cities of Pervouralsk and Karpinsk (Middle Urals) and Lermontov (North Caucasus) has proved the consistency of such approach [2,3]. At the same time, of great interest is the pooling analysis of the summarized data from all the three mentioned studies, which provides an opportunity to create a mathematical model of dependence between lung cancer development and indoor radon exposure at a wide range of observed radon concentrations.

Materials and methods

The reasons for choosing the cities of Pervouralsk and Karpinsk as study objects were as follows: the Sverdlovsk Oblast (Middle Urals) on which territory they are located is characterized by high radioactive background. By the results of the studies the average effective residential population exposure dose of radon and thoron is shown to be 1,7 mZv/year in the cities and 2,9 mZv/year in the rural regions. According to available data both cities are located in the radioactive zones and the level of cancer incidence rate was for many years higher than the average index for the region.

But the registered levels of voluminous activity of radon and thoron in the air of residential houses in these cities happened to be not high, and the range of radiation exposure of the population is narrow. Due to that the third study object has become the city of Lermontov (North Caucasus), characterized by the highest levels of radon radiation in Russia. The summarized data is shown in Table 1.

Table 1. Characteristics of persons from case and control groups included into generalized analysis

City	Number of cases	Number of controls	Total	Exposure level (WLM)
City	Number of cases	Number of controls	Total	min	Max	average
Pervouralsk	183	198	381	0.04	28.89	2.84
Karpinsk	135	181	316	0.16	77.15	13.15
Lermontov*	47	70	117	0.63	312.08	34.09
Total	365	449	814	0.04	312.08	11.5

*) all persons with occupational radon exposure excluded

Each person included in the study was characterized by a complex of 12 features, corresponding to the proved and supposed lung cancer risk factors (Table 2). Most of them do not need to be explained, while “age” and “radon exposure” need some comments. Age was registered for lung cancer patients at the moment of setting the corresponding diagnosis, for the controls – at the moment of study: 1998 – in Pervouralsk, 2001 – in Karpinsk, 2004 – in Lermontov. For estimating the radon exposure value of “working level per month” (WLM) as equal to the exposure at radon concentration 3700 Bq/m3 during 170 hours was used. According to data on the number of years (Ti), during which a person was residing at a certain place, the exposure D (WLM) was calculated:

In order to measure the indoor radon concentrations in the homes of both cases and controls the solid state nuclear track detectors LR-115, placed in special diffusion cameras, were used. Detectors were exposed in dwelling atmosphere for 2-5 months. Besides, additional measurements with application of aspiration methods and recently elaborated retrospective methods were applied. In order to estimate the indoor radon eqilibrium equivalent concentration (EEC) the equilibrium factor of 0,4 for the southern Lermontov and 0,5 for the northern Pervouralsk and Karpinsk were applied.

Table 2. Features used for multifactorial analysis

№	Feature	Feature description
1	Sex	Male, female
2	Age	Quantitative feature (years)
3	Smoking	Non-smoker; smokes less than10cig./day; 10-20 sig./day; more than 20 cig./day
4	Duration of smoking	Quantitative feature (years)
5	Alcohol use	never; seldom; once a week;more than once a week; every day
6	Duration of occupational exposure to chemical carcinogens	Quantitative feature (years)
7	Chronic lung diseases (ChrLD) in the anamnesis	absent; present
8	Intensive environmental pollution of the residential area	absent; present
9	Size of floorspace per person	Quantitative feature (m3)
10	Gas stove in the kitchen	absent; present
11	Heating of the house	central; gas; stove
12	Exposure to radon	Quantitative feature(WLM)

For multifactorial analysis mathematical methods of pattern recognition were used. The data processing consists of the following stages:

1) evaluation of sufficiency of the selected complex of features for reliable description of differences between the two classes (cases and controls);

2) quantitative estimation of informativity of each feature, which was interpreted as its influence rate;

3) evaluation of increase or a decrease of probability in cancer development under the influence of a certain feature.

The first stage was solved using the discriminant analysis methods. Some part of observations (15%) from each class was chosen for the “exam”. Based on the rest of observations the “teaching” of the computer code was conducted, in the course of which the discriminant functions (solving rules) recognizing the cases and controls, were elaborated. The criteria of its quality was the percentage of correctly recognized observations in “exam” (test) group.

To assess the informativity of each feature, which value is interpreted as the effect of risk factor in lung cancer development, a method based on estimating the difference between the average values for each feature in both classes was used. To determine the direction of influence of each feature the procedure of analysing the frequency of occurrence of the features in the studied classes was applied. All the three stages of procedure as described above was processed with the application of the software KWAZAR [2].

For correct comparison of the results of investigation with published data obtained usingthe logistic regression, all the available data were processed by the same approach as well.Of interest was to compare the results obtained in the three Russian studies with those published by other authors. But the latter have applied the methodology of logistic regression for data processing. So, for making the comparison correct and reliable we have made an additional processing of our own data with the same methodology of logistic regression.

Results

At the first stage of mathematical examination the sufficiency of 12 features for description of difference cases and controls was done. The best results of recognition at the “exam” were as follows: 91,2% for controls and 92,3% for cases. With the help of special algorithm [7] the confidence interval of the obtained recognition was calculated as 84-96%. High results obtained at the “exam” proved the initial complex containing the most significant lung cancer risk factors for the groups under study. It also has to be noted that this results were obtained using three different recognition algorithms based on potential function method, seniority committees and majority committees, that undoubtedly increases their reliability. Besides, basing on the obtained results of mathematical examination it can be noted, that the analysed cohort was representative, which is a necessary condition for receiving reliable results.

Further on the estimation of informativity of each factor was done. The obtained results are shown in Fig.1. It is seen that they practically coincide with the opinions of most of onco-epidemiologistsonthe main lung cancer risk factors. According to the obtained results most significant lung cancer risk factors are as follows: smoking, age, sex, chronic lung diseases. Non-occupational exposure to radon in the scale of informativity of features is ranked very low with its relative value equal to 0,6%.

In the course of analysis of the influence of each feature it was shown, that the risk of lung cancer is higher among males, increases depending on the intensity and duration of smoking, increases with age, is higher among persons with chronic lung diseases. Considering the residential radon exposure the weak direct relation with lung cancer risk was noted.

Fig. 1. Relative impact of various features into lung cancer development

Fig. 2. Distribution of case- and control groups depending on the value of residential exposure to radon

Judging by the obtained results it can be assumed that the based on the generalized data discriminant model, describing multifactorial nature of lung cancer development is quite reliable and noncontroversial. This conclusion allows to apply the model for estimation the risk of radon induced lung cancer per observed indoor radon concentrations. Calculations for various scenarios of exposure were done basing on the obtained model.

According to the results of Table 3, a complete elimination of indoor exposure to radon reduces the number of lung cancer patients by 1,7%, and at a limitation of the exposure to 2 WLM the reduction of patients is 1,1%. Establishing maximal exposure 20 and more WLM (corresponds to 25 years to residential radon concentration equal to 100 Bq/m3 by Russian radiation standard for new buildings) does not change the number of lung cancer patients. At the same time, increasing of minimal exposure up to 100 WLM increases the number of lung cancer patients by 5%, in case of minimal exposure of 60 WLM this number increase by 3,4%, and at 40 WLM – by 2,5%. At scenarios with minimal exposure 10 WLM and lower no increase in lung cancer incidence rate is observed. Thus, the results of modelling show that the measures aiming on decreasing the residential radon concentrations would not lead to a significant lowering of lung cancer incidence rate, because its average concentrations due to UNSCEAR data in most of the countries are not too high [ 1 ].

For the purposes of comparison, the elaborated model was applied also to various scenarios in changing the intensity and duration of smoking. According to the obtained results, in case of total cessation of smoking of all smokers, the number of lung cancer patients may be reduced by 68%. In case the smokers consume less than 10 cigarettes per day for period no longer than 10 years, such reduction may amount to 63%. At the same time, if hypothetically all the smokers and non-smokers take not less than 20 cigarettes per day for 45 years, the number of people without lung cancer may reduce by 76% (table 4.).

Table 3 . Modelling of scenarios for changing the lung cancer incidence rate depending on exposure to radon

Scenario	New value	Character of changes in lung cancer patients (cases)	Scenario	New value	Character of changes in control group
Exposure to radon is equal to 0 WLM	0	Reducing by 1,7%	Exposure to radon is equal to 0 WLM	0	No changes
Decrease of exposure higher than 2WLM	2	Reducing by 1,1%	Increase of the exposure lower than 2WLM	2	No changes
Decrease of exposure higher than 5WLM	5	Reducing by 0,6%	Increase of the exposure lower than 5WLM	5	No changes
Decrease of exposure higher than 10WLM	10	Reducing by 0,6%	Increase of the exposure lower than 10WLM	10	No changes
Decrease of exposure higher than 20WLM	20	No changes	Increase of the exposure lower than 20WLM	20	Moving to case group of 0,5%
Decrease of exposure higher than 40WLM	40	No changes	Increase of the exposure lower than 40WLM	40	Moving to case group of 2,5%
Decrease of exposure higher than 60WLM	60	No changes	Increase of the exposure lower than 60WLM	60	Moving to case group of 3,4%
Decrease of exposure higher than 100WLM	100	No changes	Increase of the exposure lower than 100WLM	100	Moving to case group of 5%

Тable 4. Modelling of scenarios for changing the lung cancer incidence rate depending on intensity and duration of smoking

Scenario	Intensity (sig./day)	Duration (years)	Character of changesin lung cancer patients (cases)	Scenario	Intensity (sig./day)	duration (years)	Character of changesin control group
Scenario	New value		Character of changesin lung cancer patients (cases)	Scenario	New value		Character of changesin control group
All the smokers do not smoke	0	0	Reducing by 68%	Increase of intensity and (or) duration of smoking for all non-smokers and those smoking less than 20 cig/day for less than 45years	More than20	45	Moving to case group of 76%
Decrease of intensity and (or)duration of smoking for those smokingmore than 10 cig/day for 10 or more years	10	10	Reducing by 63%	Increase of intensity and (or) duration of smoking for all non-smokers and those smoking less than 20 cig/day for less than 30years	более 20	30	Moving to case group of 52%
Decrease of intensity and (or)duration of smoking for those smokingmore than 10 cig/day for 30 or more years	10	30	Reducing by 24%	Increase of intensity and (or) duration of smoking for all non-smokers and those smoking less than 10 cig/day for less than 30years	20	30	Moving to case group of 30%
Decrease of intensity and (or)duration of smoking for those smokingmore than 20 cig/day for 10 or more years	20	10	Reducing by 61%	Increase of intensity and (or) duration of smoking for all non-smokers	10	30	Moving to case group of 20%
Decrease of intensity and (or)duration of smoking for those smokingmore than 20 cig/day for 30 or more years	20	30	Reducing by 19%	Increase of intensity and (or) duration of smoking for all non-smokers	10	10	Moving to case group of 2%

The results of testing show the high sensitivity of the elaborated discriminant model and particularly in the range of low exposure doses, which is of special importance. Consequently this model is proved to be a reliable technology for estimating the dose-response relationship between lung cancer and exposure to radon.

Of great interest was also the comparison of pooling analysis data based on Russian data with the results obtained in other countries. To make such comparison correct all the data was processed using the logistic regression method, but taking into account not only 2-3 risk factors applied by the other authors, but all the 12 mentioned above. Positive coefficient, connecting exposure to radon with lung cancer was obtained for the group of non-smoking males only, for the other groups it was not statistically different from 0. Evaluated 95% confidence interval for all odds ratio (OR) estimates is very wide and approaches 1 (Table 5).

Table 5.Calculation results obtained with application of logistic regression method (OR values at exposure equal to11,1 WLM, formed by 25 years long exposure to observed radon concentration of 100 Bq/m³).

sex	smoking	OR	95% confidence limit
Males	Smoker	0,92	0,78 – 1,10
Males	Non-smoker	1,8	0,78 – 4,15
females	Smoker	0,39	0,08 – 18
females	Non-smoker	0,8	-

According to calculations conducted using the regression model involved reduced set of risk factors and upper range of 95% confidence interval of ORestimated on joint material from the three Russian cities the contribution of indoor radon exposure into lung cancer incidence rate may reach 7%, that is 10 times higher than it was obtained using the discriminant multifactorial analysis.

DISCUSSION

The results of the study based on data obtained in the three Russian cities with different levels of residential radon exposure, in general give the evidence on low carcinogenic effect of radon exposure. Thus, comparing it with other cancer risk factors, the impact of radon exposure was only 0,6%, while for example, for smoking it was equal to 44%. However it has to be noted, that even at low radon exposures of almost of the persons included into the study a weak decreasing dependence between lung cancer and radon exposure value was detected. It was also proved by the results obtained by the analysis using the discriminant model, elaborated basing on the pooling data. According to this results the total absence of residential radon exposure may lead to a minor (1,7%) decrease of cancer incidence rate, while the prolonged exposure at high indoor radon concentration of 500 Bq/m³ resulted in increasing of cancer incidence rate by 5%.

At the same time, the applying the methods of logistic regression for data processing gave another results: maximum estimate of the radon contribution to carcinogenic risk in the three cities under study was 10 times higher, than it was obtained with the help of discriminant analysis. At the same time the analysis using logistic regression approach has shown the results similar to the other published results of similar studies (Table 6). Thus, the results obtained using monofactorial analysis is not substantiated enough.

Тable 6.Some published results of studies of carcinogenic risk caused by exposure to residential radon(OR values at exposure of11,1 WLM, formed by 25 years long exposure to observed radon concentration of 100 Bq/m³)

Source	sex	Smoking	OR (100 q/m³)	95% confidence limit
BEIR–VI[ 9 ]	males	Smokers	1,04	1,01 – 1,17
BEIR–VI[ 9 ]	males	non-smokers	1,11	1,02 – 1,63
Yarmoshenko et al. [17]	males and females	smokers and non-smokers	1,12	1,07 – 1,17
Darby et al [6]	males and females	smokers and non-smokers	1,084	1,03 – 1,158
Zhukovsky [18]	males	smokers	1,062	0,95 – 1,18
	males	non-smokers	1,15	0,88 – 1,44
	females	smokers	1,093	0,45 – 1,30
	females	non-smokers	1,23	0,88 – 1,72
Krewski et al [12]	males and females	smokers and non-smokers	1,11	1,00 – 1,28

One more disadvantage of the traditionally used methodological_monofactorial approach is a complexity in interpretation of the obtained results. Odds ratio (OR) itself can not be directly used for estimating risk and is usually transformed into the relative risk (RR) by the simple equating RR to OR. In its turn, the RR parameter is used to estimate the increase of lung cancer cases as compared with the incidence rate in hypothetical case of absence of residential radon exposure. The difficulty of application of RR parameter lies in the necessity for considering significant variation of background incidence rate in different countries and regions and its changes in the course of time.So it is rather difficult to interpret the OR value, whereas the results, obtained using the pattern recognition methods, clearly show the share of lung cancer cases caused by the risk factor of interest.

So the results of the analysis including the data from the three Russian cities confirm the low carcinogenic effect of indoor radon, and the mathematical model developed on this data may be used as an instrument for evaluating radon induced lung cancer risk at a wide range of exposure, including the low doses.

Literature

1. United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR), 2000. Sources and effects of ionizing radiations. Annex B. Exposures from natural radiation sources. United Nations, New York. 74 p.

2. Kazantsev VS. The KVAZAR Package for Pattern Recognition and its Applications. International Journal of Software Engineering and Knowledge Engineering 1993; V.3, N4: 439-444.

3. Lezhnin VL, Polzik EV, Kazantsev VS, et al. Evaluation of the effect of occupational and non-occupational exposure to radon on lung cancer development among the population of the city of Lermontov.Reviewof the Urals Academy of Medical Sciences2006.; № 3: 19-25 (in Russian).

4. Lezhnin VL,Polzik EV, Kazantsev VS. Advantages of Applying a Multifactorial Approach to Estimeting the Contribution of Indoor Radon to Lung Cancer Risk.Recent Advances Research Updates 2004; V. 5, № 2: 177-185.

5. Bochicchio F. Case-control studies on residential radon and lung cancer: A concise rewiew. Arch. Oncol. 2003; V.12, N1:19–24.

6. Darby S, Hill D, Deo H, Auvinen A, Barros-Dios J-M, Baysson H et al. Residential radon and lung cancer – detailed results of a collaborative analysis of individual data on 7148 persons with lung cancer and 14208 persons without lung cancer from 13 epidemiologic studies in Europe. Scand J. Work Environ Health 2006; 32 suppl.: 1-84.

7. Duda R., Hart P. Pattern Classification and Scene Analysis. A Wiley interscience publication. John Wiley&Sons. New-York-London-Sydney-Toronto;1973.

8. Howe GR, Nair RC, Newcombe HB, Abbatt JD. Lung cancer mortality (1950-80) in relation to radon daughter exposure in a cohort of workers of the Eldorado Beaverlodge uranium mine. J. Nat. Cancer Instit. 1986; V. 77: 357-362.

9. Health Effects of Exposure to Radon. Committee on Health Risks of Exposure to Radon (BEIR VI). Washington (USA). Nat. Acad. Press: 1999.

10. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. V.43. Man-made Mineral Fibres and Radon. Lyon: IARC: 1988.

11. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. V.78. Ionizing Radiation. Part 5. Some Internally Deposited Radionuclides. Lyon: IARC: 2001.

12. Krewski D, Lubin JH, Zielinski JM, Alavanja M, Catalan VS, Field RW at al. Residential Radon and Risk of Lung Cancer. A Combined Analysis of 7 North American Case-Control Studies. Epidemiology 2005; V. 16, N 2: 137–145.

13. Lubin JH, Boice JDJr Lung cancer risk from residential radon: meta-analysis of eight epidemiologic studies. J. Natl. Cancer Inst. 1997; V. 89: 49–57.

14. Lubin JH, Qiao Y.L., Taylor P.R., Yao S.X., Schtakin A., Mao B.L., et. al. Quantitative evaluation of the radon and lung cancer association in a case-control study of Chinese tin miners. Cancer Research 1990; V. 50: 174-180.

15. Pavia M, Bianco A, Pileggi C, Angelillo IF Meta-analysis of residential exposire to radon gas and lung cancer. Bull. Of the World Health Organization 2003; V. 31, N 1: 732–738.

16. Radford EP, Renard KG Lung cancer in Swedish iron miners exposed to low doses of radon daughters. New England Journal of Medicine1984;V.310: 1485-1494.

17. Yarmoshenko IV, Kirdin IA, Zhukovsky MV, Astrakhantseva SY Meta-analysis of twenty radon and lung cancer case control studies. Radioactivity in the environment (A companion series to the Journal of Environmental Radioactivity) 2005;V.7: 762-771.18. Zhukovsky MV, Yarmoshenko IV. Sex and Smoking Sensitive Model of Radon Induced Lung Cancer. IRPA Second European International Radiation Protection Association (IRPA) Congress. Proceedings of full papers CD Rom. Paris, 2006: P-049. - 11 p.

Флороценоз

Диагностика

+7 (343) 287-57-36

620075, г. Екатеринбург,ул. К. Либкнехта, 8-Б

e-mail@mail.ru

skype: 1234567

Медицина

Научно-исследовательский центр