Field calibration of the glass-based retrospective radon detectors for epidemiologic applications

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations 2008 Field calibration of the glass-based retrospective radon detectors for epidemiologic applications Kainan Sun University of Iowa Copyright 2008 Kainan Sun This dissertation is available at Iowa Research Online: Recommended Citation Sun, Kainan. "Field calibration of the glass-based retrospective radon detectors for epidemiologic applications." PhD (Doctor of Philosophy) thesis, University of Iowa, Follow this and additional works at: Part of the Environmental Public Health Commons, and the Occupational Health and Industrial Hygiene Commons

2 FIELD CALIBRATION OF THE GLASS-BASED RETROSPECTIVE RADON DETECTORS FOR EPIDEMIOLOGIC APPLICATIONS by Kainan Sun An Abstract Of a thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Occupational and Environmental Health in the Graduate College of The University of Iowa August 2008 Thesis Supervisor: Professor R.William Field

3 1 ABSTRACT The primary goal of this PhD research was to obtain critical information needed to further calibrate the novel glass-based retrospective radon detectors (RRDs) by characterizing the quantitative relationship between radon gas concentrations, the surface-deposited activities of various radon progeny, the airborne dose rate, and various residential environmental factors through both actual field measurements and Monte- Carlo simulation. Radon and radon progeny concentrations were measured, from May 2005 to May 2007, in 38 Iowa houses occupied by either smokers or nonsmokers. The investigation took into account several important indoor environmental factors, which have crucial influences on the radon progeny deposition process in homes. The long-term (3 months) surface-deposited radon progeny by species and implanted 210 Po were measured using a RRD. During the 3 months, the short-term (3-7 days) airborne radon progeny by species and bimodal size fraction were measured using a recently developed active detector. Both passive and active detectors are solid state nuclear track detectors. Airborne dose rates were calculated from unattached and attached potential alpha energy concentrations (PAECs) based on both Porstendörfer s effective dose conversion factor and that of James. Correlation analysis and multiple linear regression analysis were applied to analyze both field study results and Monte-Carlo simulation study results. Temporal and spatial variations among airborne dose rates and surface-deposited radon progeny were also investigated in actual field settings. Overall, deposited radon progeny were useful in predicting airborne dose rate in addition to the radon gas concentration. The occurrence of smoking was the most crucial environmental factor influencing the deposition process. In addition, other environmental factors were identified that served as useful covariates predicting airborne dose rates by smoking status.

4 2 The results from our current study will greatly aid the future resolution of the final reanalysis of the lung cancer risks for the Iowa Radon Lung Cancer Study (IRLCS) based on radon progeny exposure estimates obtained from RRD measurements. In addition, the results of this study will be used as the basis for a large-scale pooled analysis of the Iowa and Missouri Residential Radon Studies, both of which incorporated the use of the glassbased detectors within their study designs. Abstract Approved: Thesis Supervisor Title and Department Date

5 FIELD CALIBRATION OF THE GLASS-BASED RETROSPECTIVE RADON DETECTORS FOR EPIDEMIOLOGIC APPLICATIONS by Kainan Sun A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Occupational and Environmental Health in the Graduate College of The University of Iowa August 2008 Thesis Supervisor: Professor R. William Field

6 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL PH.D. THESIS This is to certify that the Ph.D. thesis of Kainan Sun has been approved by the Examining Committee for the thesis requirement for the Doctor of Philosophy degree in Occupational and Environmental Health at the August 2008 graduation. Thesis Committee: R. William Field, Thesis Supervisor Brian J. Smith Charles F. Lynch Daniel J. Steck Thomas M. Peters

7 To my husband Kai and our daughter Amy. ii

8 ACKNOWLEDGMENTS I would like to take this opportunity to thank all those who have contributed to this thesis, directly or indirectly. First, I would like to express my heartfelt gratitude to my advisor and the chairman of my thesis committee, Professor R. William Field, Ph.D., who introduced me to the field of radiation epidemiology and provided both outstanding guidance and encouragement during the pursuit of this degree. It has been such a fortune to have an advisor who always keeps the door open and is eagerly willing to discuss the work. I offer special thanks to Professor Daniel J. Steck, Ph.D., an authority of nuclear physics, who provided the state-of-the-art novel radon progeny detectors and helpful advice based on his advanced experimental expertise. With his help, our field investigation methods were developed and evaluated grounded in a solid foundation in physics. I also extend my sincere gratitude to Brian J. Smith, Ph.D., Charles F. Lynch, M.D., Ph.D., and Thomas M. Peters, Ph.D., for both serving on my thesis committee and imparting their wisdom during the progression of my academic work. Their helpful suggestions and advice greatly contributed to the scientific rigor of my dissertation. The cooperation of Sarah Manley in recruiting study participants and placing the detectors is greatly appreciated. The financial support from the NCI (NIH) Research Grant 5 R01 CA : Iowa and Missouri Radon Lung Cancer Studies-Phase II; and smaller grants in the form of contracts from the U.S. Environmental Protection Agency; and the Presidential Graduate Fellowship from the Graduate College of The University of Iowa, are gratefully acknowledged. iii

9 ABSTRACT The primary goal of this PhD research was to obtain critical information needed to further calibrate the novel glass-based retrospective radon detectors (RRDs) by characterizing the quantitative relationship between radon gas concentrations, the surface-deposited activities of various radon progeny, the airborne dose rate, and various residential environmental factors through both actual field measurements and Monte- Carlo simulation. Radon and radon progeny concentrations were measured, from May 2005 to May 2007, in 38 Iowa houses occupied by either smokers or nonsmokers. The investigation took into account several important indoor environmental factors, which have crucial influences on the radon progeny deposition process in homes. The long-term (3 months) surface-deposited radon progeny by species and implanted 210 Po were measured using a RRD. During the 3 months, the short-term (3-7 days) airborne radon progeny by species and bimodal size fraction were measured using a recently developed active detector. Both passive and active detectors are solid state nuclear track detectors. Airborne dose rates were calculated from unattached and attached potential alpha energy concentrations (PAECs) based on both Porstendörfer s effective dose conversion factor and that of James. Correlation analysis and multiple linear regression analysis were applied to analyze both field study results and Monte-Carlo simulation study results. Temporal and spatial variations among airborne dose rates and surface-deposited radon progeny were also investigated in actual field settings. Overall, deposited radon progeny were useful in predicting airborne dose rate in addition to the radon gas concentration. The occurrence of smoking was the most crucial environmental factor influencing the deposition process. In addition, other environmental factors were identified that served as useful covariates predicting airborne dose rates by smoking status. iv

10 The results from our current study will greatly aid the future resolution of the final reanalysis of the lung cancer risks for the Iowa Radon Lung Cancer Study (IRLCS) based on radon progeny exposure estimates obtained from RRD measurements. In addition, the results of this study will be used as the basis for a large-scale pooled analysis of the Iowa and Missouri Residential Radon Studies, both of which incorporated the use of the glassbased detectors within their study designs. v

11 TABLE OF CONTENTS LIST OF TABLES... ix LIST OF FIGURES... xi LIST OF ABBREVIATIONS... xiii CHAPTER I. INTRODUCTION...1 Background and Statement of the Problem...1 Specific Aims...5 Benefits and Significance...6 II. LITERATURE REVIEW...8 Radon and Its Decay Products...8 Health effects...8 Effective dose to lung...9 Dose conversion factor...10 Activity size distribution...11 Analysis of sensitivity...11 Room models and behavior of radon progeny...12 Methodologic Problems with Traditional Radon Exposure Assessment in Residential Radon and Lung Cancer Epidemiologic Studies...13 Random error in contemporary radon gas measurement...14 Temporal and spatial variation of residential radon concentrations...14 Missing data due to inability to accurately measure previous homes...15 Missing exposure estimates due to subject mobility...15 Measuring contemporary radon gas concentration as a surrogate for long-term dose to lung...16 An Alternative Approach to Retrospectively Reconstruct the Long-term Radon Exposure...17 Volume traps and surface traps for 210 Pb...17 Working mechanism of the glass detector...18 Attachment...19 Deposition...20 Implantation...21 Effects of aging loss...21 Systematic description of glass detector...22 Assumptions...22 Selection of surfaces...22 Detection of alpha signal from decay of 210 Po...23 Control of background alpha emission of the glass...23 Calibration of the relationship between surface implanted 210 Po activity and cumulative radon exposure...24 Laboratory calibration...24 Field calibration...24 Environmental factors that influence the calibration...24 vi

12 III. FIELD INVESTIGATION OF THE SURFACE-DEPOSITED RADON PROGENY AS A POSSIBLE PREDICTOR OF THE AIRBORNE RADON PROGENY DOSE RATE...48 Summary of Findings...48 Introduction...49 Methods...51 Site selection...51 Description of radon and radon progeny detectors...51 Retrospective radon detector (RRD)...51 Airborne radon progeny detector (ARPD)...52 Radon/thoron gas detector (GD)...52 Radon and radon progeny measurements...53 Collecting pertinent non-radiation data...53 Detectors processing, reading and results translating...54 Effective lung dose estimates...54 Data analyses...55 Results...56 Descriptive statistics...56 Correlation analysis...57 Nonsmoking rooms...57 Smoking rooms...57 Regression analyses...57 Normality check...57 Selection of deposited radon progeny...58 Selection of environmental factors...59 Discussions and Conclusions...60 IV. FIELD EVALUATION OF TEMPORAL AND SPATIAL VARIATIONS OF THE AIRBORNE RADON PROGENY DOSE RATE AND DEPOSITED RADON PROGENY...80 Summary of Findings...80 Introduction...80 Methods...81 Study design...81 Data analyses...82 Results and Discussions...83 Spatial variations...83 Temporal variations...84 Environmental factors affecting the between-room and withinroom variations...85 V. ROOM MODEL BASED MONTE-CARLO SIMULATION STUDY OF THE RELATIONSHIP BETWEEN THE AIRBORNE DOSE RATE AND THE SURFACE-DEPOSITED RADON PROGENY...95 Summary of Findings...95 Introduction...95 Methods...96 Tools and assumptions...96 Room parameters probability distributions...97 Surface-to-volume ratio of a room...97 Deposition velocities...98 vii

13 Attachment rates...98 Ventilation rate...99 Statistical analysis of MC-simulated data...99 Results Descriptive statistics Correlation analysis Nonsmoking rooms Smoking rooms Regression analyses Discussions and Conclusions VI. DISCUSSION AND FUTURE WORK APPENDIX A. FORM SENT TO THE POTENTIAL PARTICIPANTS FOR SCREENING RADON TESTS WITH E-PERMS APPENDIX B. QUESTIONNAIRE FOR COLLECTING BOTH RADIATION AND NON-RADIATION DATA IN THE SELECTED HOUSE APPENDIX C. THEORETICAL CALCULATIONS OF AIRBORNE PAEC BY BIMODAL SIZE FRACTION BASED ON ARPD TRACK DENSITY RESULTS APPENDIX D. MULTIPLE LINEAR REGRESSION EQUATIONS REFERENCES viii

14 LIST OF TABLES Table 2.1 Parameters and typical values used in the room model The typical range of equilibrium factor F and unattached fraction fp in the indoor environment obtained from the literature Major case-control studies of residential radon and lung cancer in the world Comparison of volume traps and surface traps for 210 Pb Correlation coefficients between implanted 210 Po activity and cumulative radon exposure obtained from literature Calculated effective dose conversion factors (DCF) Descriptive statistics for measured radon, airborne PAECs, dose rates, F, fp and surface deposited radon progeny by smoking status Descriptive statistics for environmental categorical variables Spearman correlation coefficients and p-values under zero-correlation none hypothesis between airborne and deposited measurements by smoking status Adjusted R-square from the natural log-scale multiple regression results for deposited radon progeny in comparison to radon Estimated relative mean change in Pdose for selected predictors using multiple regression analysis by smoking status Estimated relative mean change in Jdose for selected predictors using multiple regression analysis by smoking status The spatial coefficient of variation (COV) for short-term and long-term 222 Rn, airborne PAECs, dose rate, and deposited radon progeny by smoking status The temporal coefficient of variation (COV) for short-term and long-term 222 Rn, airborne PAECs, dose rate, and deposited radon progeny by smoking status Univariate analysis of environmental factors as fixed factors affecting variation in Pdose Univariate analysis of environmental factors as fixed factors affecting variation in Jdose Univariate analysis of environmental factors as fixed factors affecting variation in deposited 214 Po Model parameter and other input parameter probability distributions ix

15 5.2 Model constants and dose conversion factors Descriptive statistics for MC simulated radon, airborne PAECs, dose rates, F, fp and surface deposited radon progeny by smoking status Spearman correlation coefficients and p-values under zero-correlation none hypothesis between airborne and deposited variables by smoking status for MC study Model selection of log-scale MC simulated radon and deposited radon progeny as predictors of airborne dose rates (Cp, Adjust-R-squre and backward model selection method were used) Estimated relative mean (95% CI) change in Pdose for selected predictors using multiple regression analysis by smoking status for MC study Estimated relative mean (95% CI) change in Jdose for selected predictors using multiple regression analysis by smoking status for MC study C.1 Decay constants and denotations for short-lived radon progeny in ARPD C.2 Calibration factors and their uncertainties of ARPD x

16 LIST OF FIGURES Figure 2.1 Factors influencing the relationship between radon exposure and lung cancer risk Human respiratory tract Percent of particles of a specific size deposited in various parts of the human respiratory tract Effective dose per unit exposure of the thoracic regions as function of the particle diameter Compartment model of the attachment, deposition and implantation of the radon progeny The bimodal distribution showing unattached and attached activity Radon decay series Implantation by alpha recoil from radon decay products. a) Double implantation from 218 Po. b) Single implantation from 214 Po Household glass sheets analyzed with regard to implanted 210 Po normalized to the assumed cumulative radon exposure (The dashed line is the expected 210 Po activity implanted in glass assuming decay of 210 Pb being the only loss) Different calibration curves for 210 Po in glass and integrated radon exposure in smoky versus non-smoky environments The schematic drawing of RRD The schematic drawing of ARPD Photo of the radon/thoron gas detector The glass-based RRD and radon/thoron gas detector are simultaneously exposed for 90 days in a selected room A typical study room with radon/radon progeny detectors and relevant environmental factors distribution The relationship between equilibrium fraction F and unattached fraction fp Histograms of Pdose before and after log transformation Histograms of Jdose before and after log transformation The interaction effect between obstacle and deposited 214 Po in predicting airborne Pdose xi

17 4.1 Correlations between different radon measurements Correlations between repeated airborne PAEC measurements Correlations between repeated airborne dose rate measurements and surfacedeposited 214 Po measurements Schematic representation of the room model Input radon distribution of Monte Carlo model The relationship between equilibrium fraction F and unattached fraction fp under MC simulation study Sensitivity analysis for surface-deposited 214 Po C.1 The schematic drawing of the working mechanism of ARPD xii

18 LIST OF ABBREVIATIONS AML ARPD ATD BEIR CI CLL CML COV DCF EPA F fp GD HVAC ICRP IRLCS MC MRLCS-II NCI NIH PAEC RRD SSNTD UIHC acute myeloid leukemia airborne radon progeny detector alpha track detector Biological Effects of Ionizing Radiation confidence interval chronic lymphocytic leukemia chronic myelogenous leukemia coefficient of variation dose conversion factor U.S. Environmental Protection Agency equilibrium fraction of radon progeny unattached fraction of PAEC Radon and Thoron gas detector heating, ventilation and air conditioner International Commission for Radiological Protection Iowa Radon Lung Cancer Study Monte-Carlo Missouri Radon and Lung Cancer Study-Phase II National Cancer Institute National Institute of Health potential alpha energy concentration retrospective radon detector solid state nuclear track detector University of Iowa Hospitals and Clinics xiii

19 1 CHAPTER I INTRODUCTION Background and Statement of the Problem In recent years, a number of case-control epidemiologic studies have taken place in North America, Europe and China to evaluate the lung cancer risk to the general population from prolonged exposure to radon and its short-lived progeny in the indoor residential environment [1-20]. Overall, the epidemiologic studies have generally demonstrated a small, predominantly statistically non-significant, risk between residential radon gas exposure and lung cancer. The lung cancer risk posed by residential radon progeny exposure has yet to be directly evaluated. The National Research Council s Biological Effects of Ionizing Radiation (BEIR) VI Committee concluded that the apparent inconsistency in findings among residential radon case-controls studies was largely a consequence of errors in radon dosimetry [21]. While it is actually the long term exposure over past decades to radon progeny by inhalation that dominates lung doses, for a number of practical reasons it is contemporary radon gas that was measured in most of the existing radon and lung cancer epidemiologic studies. Exposure misclassification has seriously compromised the ability of these studies to detect whether or not an association exists between radon exposure and lung cancer. Case-control studies of lung cancer and domestic exposure to radon pose unique and difficult problems related to exposure assessment [22]. Field et al. have provided a summary of factors that have likely led to misclassification of radon exposure. These factors include: 1) detector measurement error; 2) failure to consider temporal and spatial radon variation within a home; 3) missing data from previously occupied homes that currently are inaccessible; 4) failure to link radon concentration with subject mobility; and 5) measuring radon gas concentration as a surrogate for radon progeny exposure [23].

20 2 Most of the published residential radon and lung cancer studies relied on a yearlong radon gas measurement using Alpha Track Detectors (ATD) as a surrogate for the actual dose to the lung from the radon progeny exposure. However, it is the radon progeny that impart the predominant dose to the lung tissues, rather than radon gas itself [21]. It is difficult to quantify how well radon gas measurements reflect radon progeny exposure from previous years. The alpha radiation that is released during the decay of two of radon s progeny ( 218 Po and 214 Po) deliver the radiologically significant dose, which increases the risk of lung cancer. The effective dose conversion coefficient for radon progeny strongly correlates with the aerosol size distribution of the progeny. The smaller particles (3 to 10 nm) provide greater exposure to the airways than radon progeny that are attached to larger aerosols (~100 nm), primarily because of their higher rate of deposition in the bronchial region. The activity size distribution of the radon progeny varies with changes in radon concentration and changes in the domestic atmosphere that include aerosol density, air movement and air exchange rate. Thus, both natural factors (e.g., weather patterns) and homeowner activities (e.g., smoking and cooking) can dramatically alter the delivered dose over short period even when the radon gas concentration keeps constant. Measuring radon gas concentrations alone could routinely introduce an uncertainty of 50% in the dose estimate [24]. Improved residential radonrelated dose estimates require measurements that depend on the actual airborne radon progeny concentrations. The potential errors and difficulties in reconstructing the historical exposure of subjects in epidemiologic studies, outlined above, have stimulated research to improve radon exposure assessment models based on statistical variation analysis, Monte Carlo methods, or additional experimental data [25-31]. In addition, these methodologic problems highlight the need for alternative approaches, which are not based on the measurement of contemporary radon gas concentration in current and past residences [32].

21 3 In the late 1980s, it was proposed that the measurement of surface implanted 210 Po could be used as a retrospective monitor for radon exposure [33, 34]. Since then, the techniques and field methods for accurately measuring the surface implanted 210 Po and reconstructing past radon levels in the air from the surface implanted 210 Po have been investigated by numerous scientists in both Europe and the United States [35-61]. However, the major research challenge of this technique in reconstructing past airborne radon progeny levels from the measured surface implanted 210 Po is effectively controlling for the potentially influential environmental factors retrospectively. Previous epidemiologic or investigational studies that performed both glass implanted 210 Po measurements and traditional contemporary radon gas measurements tended to make a simple linear regression to find an unadjusted relationship between implanted 210 Po activity and cumulative radon exposure (contemporary radon gas concentration multiplied by exposure duration), or at most stratify the relation by smoking status [48, 53, 56, 59]. To improve radon progeny dose estimates, the Iowa Radon and Lung Cancer Study (IRLCS) and Missouri Radon and Lung Cancer Study-Phase II (MRLCS-II) utilized novel retrospective radon detectors for reconstructing past radon progeny concentrations by analyzing the alpha activity deposited on and implanted in glass surfaces [36, 37, 62-64]. The surface-deposited activities of the various radon progeny reflect both the airborne radon progeny activities and atmospheric conditions. The IRLCS s Retrospective Reconstruction Detector (RRD) [36, 62, 63] incorporates three track registration chips to simultaneously measure contemporary radon gas, contemporary surface-deposited radon progeny, and surface-implanted 210 Pb through its alpha emitting progeny, 210 Po. Theoretically, contemporary airborne radon progeny concentrations can also be estimated through an empirical model using the direct measurements of the short-lived radon progeny that are deposited on surfaces and the radon gas concentration. However, realistically, significant uncertainties exist in the

22 4 interpretation of the measured contemporary surface-deposited radon progeny and surface-implanted 210 Po, because the major room-specific atmospheric factors such as aerosol density and air movement affect radon progeny deposition in houses. Precisely, because of the uncertainty in the interpretation, the IRLCS has not moved forward performing risk analyses based on the glass-based RRD measurements. The IRLCS glass-based RRD was just developed at the time the IRLCS initiated and the IRLCS investigators thought it would be cost effective to include them into the existing study (prior to calibration) rather than attempting to initiate another large scale case-control study at some later time incorporating the calibrated RRDs. This dissertation work is supported by research grant 5 R01 CA (January 1, Present) from National Cancer Institute (NCI), NIH: Iowa and Missouri Radon Lung Cancer Studies: Phase II and smaller grants in the form of contracts from the U.S. Environmental Protection Agency (EPA). The overall goal of the initial NCI funding was to further refine the estimated lung cancer risk posed by residential radon decay product exposure using a novel glass-based retrospective radon progeny reconstruction detector. This grant has three major components: 1) Field calibration and laboratory validation of the glass-based retrospective radon detectors; 2) analysis of the risk estimates from the Iowa Radon Lung Cancer Study incorporating radon progeny exposure estimates obtained from RRD measurements, rather than radon gas; 3) calculation of risk estimates from a pooled analyses of retrospective radon detectors exposure results for the IRLCS and the MRLCS-II. The proposed activities in this grant include calibration of the glass-based detectors in both a controlled exposure room, which simulates various residential depositional environments, and in actual field settings, where the environmental conditions will vary in an uncontrolled fashion. Once the glass-based detectors are calibrated as proposed, a reanalysis of the lung cancer risk factors for the IRLCS, based on radon progeny exposure estimates obtained from RRD measurements, can be performed.

23 5 Specific Aims The specific aims enumerated below will answer, or aid in the future resolution of the following questions. Question 1) Do the RRDs provide consistent and reliable results in the actual domestic environment? Question 2) What is the quantitative relationship between the airborne radon progeny dose rate and the alpha activity from surface-deposited radon progeny under different aerosol and ventilation situations in home environment? Is the surfacedeposited radon progeny activity a good predictor for airborne radon progeny dose rate? Question 3) What is the quantitative relationship between the airborne radon progeny dose rate and the radon gas concentration under different aerosol and ventilation situations in the home environment? Is the radon gas concentration a good predictor for airborne radon progeny dose rate? Question 4) What is the quantitative relationship between the airborne radon progeny dose rate and the implanted 210 Po activity under different aerosol and ventilation situations in home environment? Is the implanted 210 Po activity a good predictor for airborne radon progeny dose rate? Question 5) Do all the relationships mentioned above vary temporally? If so, how do they vary and what environmental factors affect the variability?

24 6 Question 6) Do the above relationships found in the actual field settings agree with those found from a Monte Carlo-based mathematical room model commencing with that of Jocobi, extended by Porstendörfer and other scientists? There are three specific aims for this dissertation. The specific aims will be accomplished by applying standard techniques in radon/radon progeny measurements, environmental monitoring, Monte Carlo model-building, statistical data analysis, and results interpretation. Aim 1) Field calibration of the RRD surface-deposited radon progeny as a predictor of the airborne radon progeny dose rate. Aim 2) Field evaluation of temporal and spatial variations of surface-deposited radon progeny and the airborne radon progeny dose rate. Aim 3) Comparison of the field findings with Monte Carlo model calculations. Benefits and Significance Comprehensive investigations of the relationship between airborne radon progeny concentrations and the measured surface-deposited radon progeny concentrations, considering multiple influential environmental factors such as ventilation and different sources of indoor aerosol and their activity size distributions, are extremely rare in both laboratory settings and actual field settings. Therefore, our laboratory and field investigation of the performance and variations of the novel glass-based RRD, simultaneously using several other novel alpha track detectors like the airborne radon progeny detector (ARPD) and radon/thoron gas detector, will benefit both scientific

25 7 knowledge and physical technology regarding the radon dosimetry. We anticipate these findings will lead to better estimation of lung cancer risk posed by cumulative residential radon progeny exposure. The results from our current study will aid in the future resolution of the final reanalysis of the lung cancer risks for the IRLCS based on radon progeny exposure estimates obtained from RRD measurements. In addition, the results of this study will be used as the basis for a large-scale pooled analysis of the Iowa and Missouri Residential Radon Studies, both of which incorporated the use of the glass-based detectors within their study design.

26 8 CHAPTER II LITERATURE REVIEW Radon and Its Decay Products The health aspects and general behavior of radon and its decay products have attracted the increased attention of researchers worldwide since the 1970s. For detailed information on the characteristics and behavior of radon and its decay products in indoor air, it is the best to turn to the comprehensive literature review on this subject [65-72]. Health effects The average individual in the United States receives more radiation dose from exposure to indoor radon decay products than from any other source of natural or manmade radiation [73]. Occupational exposure to radon in uranium and other mines is a well documented cause of lung cancer [21, 74-80]. In addition, there is now direct evidence that prolonged radon exposure in homes represents a significant health risk. Recently, two large-scale epidemiologic studies, a North American study [81, 82] and a European study [83, 84], pooled data from 20 previously performed epidemiologic studies that directly assessed the lung cancer risk of prolonged residential radon exposure. Both the North American as well as the European pooled studies support the risk projections extrapolated from occupational studies of radon-exposed underground miners [21], and provide direct evidence that prolonged residential radon exposure represents a major cause, even below the U.S. EPA s action level of 4 pci/l, of lung cancer. Empirical studies performed by Field et al. [85] suggest that pooled risk estimates are likely underestimates of the true risks. Nonetheless, based predominantly on the studies of radon-exposed miners, the EPA estimates that approximately 21,000 radon-related lung cancer deaths occur each year in the United States [86] making it one of the most significant public health risks in the United States [87].

27 9 A few ecological studies and case-control studies have indicated that exposure to indoor radon could be of some importance as a cause of other tumors, especially acute myeloid leukemia (AML), melanoma and kidney cancer, but the studies were not generally consistent with each other and most of them found no significant associations [88-102]. However, a recent case-cohort study by Řeřicha and colleagues [103] examined the incidence of leukemia, lymphoma, and multiple myeloma in Czech uranium miners. The investigators noted that exposure to radon and its progeny was positively associated with an increased risk of developing leukemia. A surprising result of the research was that chronic lymphocytic leukemia (CLL), which was not previously believed to be radiogenic, was also positively associated with radon exposure. In addition, both myeloid leukemia and Hodgkin lymphoma were also positively associated with radon exposure. The latter findings regarding myeloid leukemia and Hodgkin lymphoma were not statistically significant. Another recent study performed by Smith et al. [102], a Bayesian analysis that allowed for the joint prediction of county-average radon levels and estimation of the associated leukemia risk, reported suggestive evidence of a positive association between county radon levels and both CLL and chronic myelogenous leukemia (CML). Effective dose to lung Most of the published residential radon and lung cancer studies relied on the radon gas measurement as a surrogate for the actual dose to the lung from radon progeny exposure. Inhalation of the short-lived radon progeny ( 218 Po, 214 Pb, 214 Bi/ 214 Po) in homes, in outdoor atmosphere, and at work places yields the greatest amount of the natural radiation exposure to man [104]. However, there are numerous factors influencing the relationship between radon progeny exposure and actual dose to lung, as well as the relationship between dose and response of lung (Figure 2.1).

28 10 While the implanted 210 Po activity in glass provides a possible way to retrospectively reconstruct the historical short-lived radon progeny, we need to investigate how different environmental factors influence the dose imparted to the lung by the radon progeny so that we can obtain estimates of the historical cumulative effective dose from the cumulative radon exposure. Dose conversion factor The dose conversion factor (DCF), which provides the relationship between effective dose and potential alpha energy concentration (PAEC) of inhaled short-lived radon progeny, is calculated with a dosimetric approach [105]. The calculations are based on a lung dose model from the International Commission for Radiological Protection (ICRP 66) [106, 107]. The activity size distribution, in term of PAEC, is an important input parameter in DCF calculations. An illustration of the human respiratory tract is shown in Figure 2.2. The deposition efficiency, as a function of aerosol particle diameter in various parts of human respiratory tract, is shown in Figure 2.3 and the effective dose per exposure unit of the bronchial (BB), bronchiolar (bb), and alveolar region (AI), as a function of aerosol particle diameter calculated for short-lived radon progeny, is shown in Figure 2.4. The deposition of radon decay products differs in each of the respiratory regions [108]. From Figure 2.4, we can see that the bronchial region receives a dominating proportion of the total lung dose in the size range of cluster mode (<5 nm), while the bronchiolar region receives the highest dose in the size range of the aerosol attached mode (>10 nm). Even in the size range from nm, where the alveolar region has relatively higher particle deposition efficiency, the effective dose contribution of this region is more than five times lower than the contribution of the bronchial and bronchiolar region. In addition, 90-95% of malignant primary lung tumors are bronchogenic. Taking into account the data regarding cancer of the human respiratory

29 11 tract, Porstendörfer set the relative sensitivity between bronchial (WBB), bronchiolar (Wbb) and alveolar-interstitial (WAI) lung region as WBB:Wbb:WAI = 0.80:0.15:0.05. This partitioning of detriment leads to higher effective doses by the decay product clusters and lower values by the aerosol fraction than the partitioning WBB:Wbb:WAI = 0.33:0.33:0.33 recommended by ICRP 66 and revisited by James [79, 104]. Activity size distribution The relative activity size distribution can be divided into: 1) relative size distribution of the unattached radon progeny, 2) relative size distribution of the radon progeny aerosol, and 3) the unattached fraction in terms of PAEC [104]. The dominant parameter which influences the unattached fraction is the attachment rate, which depends on the number concentration of the aerosol Z. The unattached fraction fp can be estimated by the following equation: 414 Zcm ( ) f prn, 3 The activity size distribution in the atmosphere is generated in two steps. First, the freshly generated radionuclides from the decay of radon gas rapidly react with trace gases and air vapors and become small clusters, called unattached progeny with diameters between 0.5 to 5 nm. Besides the cluster formation, the progeny attach to the existing aerosol particles in the atmosphere generating radioactive aerosols with the primary maximum size distribution of nm [109]. (2.1) Analysis of sensitivity An important aspect in model uncertainty analysis is the evaluation of input parameter sensitivities with respect to model outcomes [110]. Marsh and others have performed parameter uncertainty analyses based on the ICRP 66 human respiratory tract model to derive the probability distribution of the weighted equivalent dose to lung for an adult per unit exposure to radon progeny in the home [111, 112]. Mohamed [113]

30 12 assessed the influence of biological and aerosol parameters on human lung dose based on measurements of the activity size distribution of indoor radon progeny, their unattached fraction, and potential alpha energy concentration. The radioactive aerosol and biological parameters are varied in order to assess the DCF uncertainty arising from the uncertainties of these parameters. The main emphasis is on the variation of the ventilation rate, breathing mode, critical cells for the induction of lung cancer, and the parameters of the attached and unattached activity size distribution of the radon progeny. Room models and behavior of radon progeny The concentrations of radon and its decay products are influenced by the complex interaction of a number of processes, the most important of which are radioactive decay (especially α-decay), ventilation, attachment to aerosols, and deposition on surfaces [65, 114]. Radon progeny are divided into four different groups depending on their various states: attached, unattached, deposited, and implanted. The transition between these compartments is described by the room model parameters (Figure 2.5). A number of mathematical models dealing with indoor radon progeny have appeared in recent decades, commencing with that of Jocobi [115], which has been extended by Porstendörfer et al. and others [ ]. The important parameters and typical values for average indoor conditions used in modeling the behavior of radon progeny are shown in Table 2.1. Assessing radon progeny and their size distribution, equilibrium fraction, and unattached fraction has been an active area of research in recent years [ ]. To reasonably estimate the population dose from radon; equilibrium factor, F, and unattached fraction, fp, of radon progeny are important parameters [127]. Both F and fp values follow a lognormal distribution. F and fp values are inversely correlated. The aerosol properties, followed by the ventilation rate, are the most important parameters influencing F and fp. High F values are correlated with high aerosol particle concentrations and low ventilation rates. Table 2.2 summarizes several measured ranges

31 13 or arthritic means of equilibrium factor F and unattached fraction fp in indoor environments obtained from the literature. Another good summary of F and fp values in indoor atmospheres, under actual living conditions, can be found in Porstendörfer s review paper [66]. Methodologic Problems with Traditional Radon Exposure Assessment in Residential Radon and Lung Cancer Epidemiologic Studies In recent years, a number of large population-based epidemiologic studies have taken place in North America, Europe, and China. Pooled analyses of these studies are either finished or in progress to evaluate the lung cancer risk to the general population from exposure to radon and its short-lived progeny in the indoor residential environment [1-20, 81, 84, ]. While it is actually the long term exposure to radon progeny by inhalation that dominates lung doses, for a number of practical reasons it is contemporary radon gas that is measured in most of these studies. Exposure misclassification has seriously compromised the ability of these studies to detect whether an association exists between radon exposure and lung cancer. Case-control studies of lung cancer and domestic exposure to radon pose unique and difficult problems related to exposure assessment [22]. Field and Steck have noted that misclassification of radon exposure has arisen primarily from: 1) detector measurement error; 2) failure to consider temporal and spatial radon variation within a home; 3) missing data from previously occupied homes that currently are inaccessible; 4) failure to link radon concentrations with subject mobility; and 5) measuring radon gas concentration as a surrogate for radon progeny exposure [23]. The major epidemiologic studies of residential radon and lung cancer are summarized in Table 2.3 with special emphasis on radon exposure assessment.

32 14 Random error in contemporary radon gas measurement Due to the availability of reliable radon measurement techniques, errors in this category are perhaps the smallest and least variable [137]. The passive alpha track detector (ATD) is the primary type of radon detector that has been used for radon measurements in epidemiologic studies. If well calibrated ATDs are used, a measurement error of less than 15% is easily achievable at 4 pci/l. Unfortunately, few epidemiologic studies adequately assess, document, and report their detectors accuracy and precision. However, the Iowa Radon Lung Cancer Study (IRLCS) investigators have described their quality assurance and quality control (QA\QC) procedures as well as the results of QA\QC testing [138]. Temporal and spatial variation of residential radon concentrations The temporal variation of residential radon concentrations can be accounted for, without having to adjust for seasonal radon variation to a certain degree, by the use of yearlong ATD measurements. Nonetheless, year-to-year home radon variation may be significant [ ]. While a large percent of recent case-control epidemiologic studies performed yearlong radon measurements, only a few of them accounted for spatial radon variations between and within the levels of a home. To calculate overall exposure, most studies measured only one or two rooms (usually the living room or bedroom or both) in each home and simply took the average of the radon concentrations. Preliminary findings of the IRLCS have noted radon concentrations differing by a factor of 20 in different areas of the same home [23]. The wide individual variations noted in radon concentrations serve as a reminder of the importance of performing multiple radon measurements in various parts of the home when estimating home radon concentrations [145, 146].

33 15 Missing data due to inability to accurately measure previous homes Most residential studies tried to obtain radon measurements for every dwelling occupied by the study participant over the previous 30 years, but gaps in the participants exposure history often occur because it may not be possible to access and measure all residences. For example, in a study in Winnipeg, Canada [2], each subject had lived in nine residences on average of which five were in the study area of Winnipeg. Further, on average, only 3 of the Winnipeg homes were monitored for radon. The previous homes may be out of the sampling area or not within the measurement protocol specifications, nonexistent, unoccupied, or the current owner may have refused to allow radon testing [22]. These gaps seriously decrease the studies statistical power to reveal an association. A few studies, such as IRLCS [5], included only subjects who had occupied their current home for at least 20 years to avoid gaps in the exposure window. In addition, the contemporary annual average radon level in a dwelling may differ considerably from that in the past due to both building modifications and differences in the usage and ventilation preferences of the present as compared to previous occupants. Thus, even if all previous residences for each subject still exist and permission to measure radon in the homes is obtained from the present owners, the cumulative radon exposure estimated on this basis may differ considerably from the actual cumulative exposure the subject actually received [137]. Missing exposure estimates due to subject mobility To date, most epidemiologic studies have relied on radon measurements in one to two rooms to characterize the entire domestic exposure, without having attempted to link the subjects temporal and spatial home occupancy patterns with multiple radon measurements within a home. Nor have studies linked the subjects mobility outside the home with the measured or estimated radon exposure outside the home. These linkages

34 16 allow calculation of retrospective cumulative radon exposure over a given time period [23]. Only the IRLCS performed comprehensive retrospective radon exposure assessment, which not only measured on-site radon levels in multiple areas of the home, but linked the historic subject mobility with radon concentrations in the residence, outdoor, and other buildings [147]. The comparative study by Field et al. [148] of different exposure scenarios showed that the power of an epidemiologic study to detect an excess risk from residential radon exposure is enhanced by linking spatially disparate radon concentrations with the subject s retrospective mobility. Measuring contemporary radon gas concentration as a surrogate for long-term dose to lung Most of the published residential radon and lung cancer studies relied on a yearlong ATD radon gas measurement as a surrogate for the actual dose to lung from radon progeny exposure. However, it is radon progeny that deliver the radiologically significant dose to the lung tissues, rather than radon gas itself [21]. It is unknown how well these radon gas measurements reflect radon progeny exposure from previous years. Radon concentration alone may not be an adequate surrogate to measure for lung cancer risk in all residential radon epidemiologic lung cancer studies [32, 149, 150]. The alpha radiation that is released during the decay of two of radon s progeny ( 218 Po and 214 Po) delivers the majority of the dose, which increases the risk of lung cancer. The effective dose conversion coefficient for radon progeny strongly correlates with the aerosol size distribution of the progeny. The smaller particles (3 to 10 nm) provide greater exposure to the airways than radon progeny that are attached to larger aerosols (~100 nm), primarily because of their higher rate of deposition in the bronchial region. The radon progeny activity size distribution varies with changes in radon concentration and changes in the domestic atmosphere that include aerosol density, air movement, and air exchange rate. Thus, both natural factors (e.g., weather patterns) and homeowner behaviors and

35 17 activities (e.g., smoking, opening windows) can dramatically alter the delivered dose over short period even when the radon gas concentration is constant [151, 152]. An Alternative Approach to Retrospectively Reconstruct the Long-term Radon Exposure The potential errors and difficulties outlined above in reconstructing the historical exposure of subjects in epidemiologic studies have stimulated a great deal of research aiming to refine or improve the radon exposure assessment based on statistical variation analysis, regression calibration method, Monte Carlo methods, or additional experimental data [25-31, ]. In addition, these methodological problems highlight the need for alternative approaches, which are not based on the measurement of contemporary radon in present and previous residences. Volume traps and surface traps for 210 Pb The measurement of the long-lived radon progeny, starting with 210 Pb (T 1/2 =22.3 y), and remaining in the dwelling provides an alternative for measuring cumulative, past indoor radon exposure. The alpha particles emitted from 210 Po, a 210 Pb decay product, provides the easiest measure of the trapped activity. Two different techniques are presently being investigated. In the volume trap technique, the 222 Rn diffuses into a porous material, such as mattresses and soft furnishings found in all dwelling, where the radon progeny decayed from 222 Rn inside the material will remain and accumulate. Knowing the age of the porous material and volume concentration of 210 Po, provides an estimate of 222 Rn exposure to the material [50, 156, 157]. In the surface trap technique, a fraction of airborne short-lived radon decay products deposit on room surfaces, such as glass, and some of this surface activity is permanently implanted to a maximum depth of approximately 100 nm into the surface of the glass as a result of alpha recoil following the decays of 218 Po and 214 Po [158]. The

36 18 activity from the buildup of 210 Pb in the glass and its descendant, the alpha emitter 210 Po, constitute a surface trap. The comparison of volume traps and surface traps is shown in Table 2.4. The feasibility of using long-lived radon decay products as radon retrospective monitors was discussed by Samuelsson and Johansson and the conclusion was reached that only surface implantation sinks are potentially well suited for long-term studies at radon level met in dwellings [39]. The practical use of volume traps in epidemiologic studies is somewhat limited, because most homeowners would not allow removal of a section of the volume trap (e.g., mattress). Therefore, in the following review, we will focus on the surface traps and glass-based detectors. Working mechanism of the glass detector The naturally implanted atoms of radon progeny constitute a memory of past air concentration of short-lived radon progeny. 210 Pb has a long half life (22y) and therefore 210 Pb and the following decay product, alpha emitter 210 Po, reflect the radon history several decades back in time. The measurement of implanted, long-lived radon progeny is very attractive to radioepidemiologist since the timescale for the build-up of lung cancer risks and 210 Pb activity are assumed to be similar [35]. In the late 1980s, it was first proposed that the measurement of surface implanted 210 Po could be used as a retrospective monitor for radon exposure [33, 34]. Since then, the techniques and field methods for accurately measuring the surface implanted 210 Po and reconstructing past radon concentrations in the air from the surface implanted 210 Po have been investigated by several scientists in Europe and the U.S. [35-61, 63, ]. There are two physical processes that are fundamental to retrospective radon monitoring by surface traps : (1) plate-out or deposition of short-lived radon progeny and (2) implantation of nuclides by alpha recoils. The first process, plate-out, is the necessary marriage between the radon progeny and the microscopic surface. Sticking to a surface

37 19 prevents the radon progeny from leaving the indoor environment by ventilation. The second process, implantation, is the process that a daughter nucleus recoils into the outermost layer of a surface (e.g. glass surface) from an outgoing alpha particle. Implantation prevents the subsequent radioactivity from the interference of human beings, who sooner or later will disturb the surface (e.g., cleaning) [35]. Attachment Indoor air normally contains copious numbers of aerosol particles, ranging in size from almost atomic dimensions up to several micrometers in diameter. Cooking and smoking are prominent sources of indoor aerosols, but there are many other sources as well (e.g., candles). Aerosols play a crucial role in the behavior of radon progeny, like determining the concentrations of radon progeny in the air and thus potentially influencing the radiation dose from radon progeny. The progeny can coarsely be classified as being in unattached or attached mode (Figure 2.6). The bimodal mode is the effect of the attachment process of radon progeny to the aerosols. Most of the unattached activity becomes attached to ambient airborne particles within seconds. The degree of attachment to the local aerosol depends on the aerosol concentration. A high particle concentration is associated with a high attached fraction [119]. Unattached and attached activities behave differently and therefore it is of interest to separate these two particle modes. Particle size separation of unattached and attached can be achieved by employing the wire screen technique [109], which principally exploits the higher Brownian motion of smaller particles to collect them, while larger particles pass through [114]. The collection efficiency depends on wire screen parameters, flow rate and particle size. Due to changes in indoor living conditions (e.g., heating, cooking, cleaning, lifestyle, smoking habits, etc.) over recent decades, aerosol concentrations and radon

38 20 progeny concentrations, as well as their activity size distributions, likely would have exhibited significant variations even in cases where the radon concentrations may have remained relatively constant. Deposition The interaction between aerosols and indoor surface are usually presented by the deposition rate shown as the following equations: S λd = Vdep V V = J / C dep λ deposition rate, s V d dep deposition velocity, ms S surface to volume ratio, m V -2-1 J net flux of aerosols, gm s C aerosol concentration, gm (2.2) It is generally assumed that the deposition velocity for attached activity is about 100 times lower than for the unattached activity [66]. The retrospective determination of radon concentrations in dwellings by means of the measurement of the 210 Po surface activity is subject to various uncertainties. These result partly from the values assumed for the equilibrium factor F and for the unattached fraction fp, and, more importantly, from differences in the deposition velocities of short-lived decay products of 222 Rn, caused by varying conditions of turbulence. The deposition velocities for 218 Po and 214 Pb unattached fractions differ significantly, with average ratio 4:1, subject to the inferior measurement precision for the deposition velocity of 214 Pb [163]. Schmidt investigated the surface deposition of short-term radon progeny for his doctoral dissertation [163, 164], while Roos presented a comprehensive summary of deposition velocities of unattached and attached activity, from literature, for her doctoral dissertation [114].

39 21 Implantation The decay scheme of radon (Figure 2.7) shows that the decay of only two nuclides can generate alpha recoils that lead to implanted 210 Po activity. These are the alpha decays of 218 Po and 214 Po. The two occurrences are of prime interest for the buildup of 210 Po activity in the glass surface (Figure 2.8) [60, 114]. The implantation probability depends on particle size if the deposited activities are in both unattached and attached forms. Since the range of recoiling nuclides from alpha decay is short, approximately 50 nm, the dust on the surface may also prevent radon progeny from implanting in the glass [38]. The effects of hydration layers are also postulated to be a source of implantation uncertainties [54]. Effects of aging loss With the simplifying assumption that the environmental conditions (ventilation and air turbulence) have remained approximately constant over the years, the implanted 210 Po can be assumed to be proportional to the cumulative radon concentration. However, a discrepancy between radon exposure and glass activity was found in model calculations and measurements, especially concerning older glass sheets [49, 54, 165]. Roos reported the activity lost from the glass samples with an effective half-life of about 10 y [114]. The apparent loss of 210 Po activity (Figure 2.9) could be due to the corrosion of the glass [166]. The corrosion rate would of course influence the glass as an integrating radon monitor. However, a field study of Steck et al. in 990 Iowa houses and 24 Minnesota houses failed to confirm the above aging loss effects. In fact, no significant reduction in the 210 Po activity-cumulative radon exposure ratio was observed in the study homes [162].

40 22 Systematic description of glass detector Assumptions The use of glass as a retrospective radon monitor is based on the following key assumptions [36]: 1) The product of the average radon concentration and the integrated exposure time are linearly related, over a wide range of exposure, to the activity of 210 Po implanted in a glass surface. 2) Theoretical calculations of exposure versus implanted activity agree with both experimental chamber results and with measurements on household glass. 3) The implanted fraction of 210 Po is not significantly affected by domestic cleaning or thin films. Deposited activity can vary widely, thus the determination of radon histories should rely on implanted activity. 4) Measured glass activity provides a better estimate of long-term radon exposure than short-term radon measurements. The effectiveness and general applicability of the method is improved by selecting a (radon to 210 Pb) regression model that reflects the deposition environment of the glass. 5) The implantation depth is shallow, but deep enough to resist removal by normal domestic cleaning methods. Selection of surfaces In order to use glass as a retrospective monitor, it has been suggested that the following criteria be fulfilled [47]: 1) The glass surface must be ordinary, smooth glass without visible coatings or colorings (not lead crystal). 2) The glass should be vertically mounted and facing the interior of the room. 3) The glass surface must have a known age, greater than 10 years. The age must be accurate to within 10%, if possible.

41 23 4) The glass surface should not have been exposed to a significant radon environment prior to the exposure site of interest. 5) The glass surface should be large enough to collect sufficient total activity. 6) The glass surface must be free to accumulate the activity, which means that the glass was not stored in a cabinet or drawer for long periods of time and no large objects should be in front of the glass surface (>25cm). 7) The glass surface should not be affected by strong air currents, such as those produced by fans or open windows. Detection of alpha signal from decay of 210 Po To date, most studies on 210 Po activity implanted in the glass surface use solidstate nuclear track detectors (SSNTD) (especially CR-39) [36, 37, 41, 42, 45, 47, 48, 53, 59, 157, 162, 167, 168]. A very detailed description of various aspects of track detection can be found in a recent review article of D. Nikezic and K.N. Yu [169]. Samuelsson etc. [34, 35, 51] performed 210 Po detection by large area pulse ionization chambers with higher efficiency and accuracy, but not suitable for large scale field studies. A new and fast radiochemical method has been developed to investigate surface beta activity concentration of 210 Bi by Fehér etc [165, 170]. A hand-held alpha ZnS scintillation probe was recently successfully used to detect 210 Po on glass surfaces by von Philipsborn and Just [171]. Control of background alpha emission of the glass Due to the naturally distributed uranium and thorium series in the glass matrix, the intrinsic alpha activity contamination in glass must be controlled [55]. In a Minnesota study, the unexposed glass samples exhibited an contamination rate equivalent to 1.3±1.0 Bqm -2 of 210 Po [162]. Steck et al. placed a metalized Mylar cover between the CR-39 detector and the glass to reduce UV damage and tracks from the natural alpha

42 24 emitters present in the glass [59]. Other researchers have used a LR115 companion detector to measure intrinsic activity [53]. Calibration of the relationship between surface implanted 210 Po activity and cumulative radon exposure Contemporary measurements of radon concentration in the dwellings of individuals, in a case-control study, are traditionally used as a surrogate to assess past radon exposure history. The results are then time-weighted using the length of occupancy in each home to obtain a cumulative estimate of past exposure usually in unit of kbqm -3 y. Therefore, in order to interpret the results from the implanted 210 Po- 210 Pb activities (in Bqm -2 ), it is of interest to calibrate the relationship between surface 210 Po activities and cumulative radon exposure of glass surfaces. Laboratory calibration Under laboratory conditions, very good correlation exists between radon exposure and implanted 210 Po surface activity. Lively and Steck performed calibration of the glass detectors in four radon chambers with radon concentration varying from 100 Bqm -3 to 9 MBqm -3. The linear regression correlation coefficients (R 2 ) for the radon chamber data are all >0.9 [36]. Field calibration The correlation coefficients between implanted 210 Po activity and cumulative radon exposure obtained from field data were much lower than those noted from the laboratory and varied in a large range (Table 2.5). Environmental factors that influence the calibration Table 2.5 shows that the correlation coefficients obtained from the actual field studies are much lower than those obtained from laboratory studies. Study results indicate no significant difference in the activity-exposure ratio based on observable films

43 25 [162], and different cleaning habits only marginally interfere with the implantation of recoils into the glass matrix [35]. Therefore the scattered results are probably due to different progeny deposition environments among houses, leading to multiple exposureversus-activity ratios. For example, in the Iowa study, the correlation R 2 improved from 0.5 to 0.7 when the measured deposited surface activities are added to the linear regression model [63]. The major environmental factors influencing the deposition of radon progeny include: deposition velocity, ventilation rate, aerosol conditions, and the surface to volume ratio [56]. Aerosol conditions play a crucial role in the calibration, since unattached and attached fraction of radon progeny have quite different deposition velocities. This would be expected to result in a different effect on dosimetry; where the calibration curve relating the actual historical radon exposure to the remaining 210 Po activity in the glass would be different, as shown in Figure 2.10, in historically smoky and non-smoky environments [40].

44 26 Table 2.1 Parameters and typical values used in the room model Symbol Meaning Typical Value F Equilibrium factor between radon progeny and radon Low aerosol concentration: 0.2±0.1 High aerosol concentration: 0.8 Z Aerosol concentration cm -3 f p Unattached fraction of PAEC u v d u λ d a λ d Deposition velocity for unattached progeny 5-10 mh -1 Deposition rate for unattached progeny h -1 Deposition rate for attached progeny h -1 λv Ventilation rate h -1 λa Attachment rate h -1 p 1 Recoil factor 0.83 S/ V Surface to volume ratio 1-25 m -1

45 27 Table 2.2 The typical range of equilibrium factor F and unattached fraction fp in the indoor environment obtained from the literature Author Clouvas, A. et al. [172] Chen, C. J. et al. [127] Mohamed, A. [173] El-Hussein, A. [174] Ortega, X. et al. [175] No. of dwellings ±0.10 ( ) ( ) F fp Comments ( ) Measured with a SILENA (model 4s) instrument fp measured with SARAD EQF ± ±0.02 Grab sampling ± ±0.01 Realistic living rooms with low ventilation rates 4 ( ) (0-0.70) Continuous measurements of radon and progeny George et al. [65] Porstendörfer et al. Several 0.3±0.1 ( ) [65] Tokonami, S et al. [176] Reineking, A. et al. [177] Laboratory Farid, S. M. [178] ±0.23 ( ) <0.005 Stable condition Air circulation system on without additional aerosol with additional aerosol Measured with integrated SSTND

46 Table 2.3 Major case-control studies of residential radon and lung cancer in the world (Part I) Study North America Enrollment Period No. of Cases No. of Controls Sex a Alive Cases Proportion (%) Mean Radon (Bq/m 3 ) Excess OR b (95% CI c ) at 100 Bq/m 3 New Jersey-I+II [1] F 57 (53% for controls) (-0.28, 0.97) Missouri-I [3] F Missouri-II [4] F Iowa [5] F ATD d : 0.04 (-0.13, 0.57) RRD e : 0.63 (0.07, 1.93) All subjects: 0.16 (-0.03, 0.61) Living subjects: 0.33 (0.02, 1.23) Connecticut [6, 135] M+F (-0.21, 0.51) Utah-South Idaho [6, 135] M+F (-0.20, 0.55) Winnipeg (Canada) [2] M+F 65 (89% for controls) (-0.14, 0.05) Europe Sweden: Stockholm [7] F (-0.05, 1.54) Sweden: nationwide [8] M+F 10 (91% for first control group, 10% for the second control group) (0.01, 0.22) Adjusted for measurement error (COV f =50%): 0.17 (0.03, 0.37) 28

47 Table 2.3 (Part I continued) Study Enrollment Period No. of Cases No. of Controls Sex a Sweden: nationwide [8] M+F Alive Cases Proportion (%) 10 (91% for first control group, 10% for the second control group) Mean Radon (Bq/m 3 ) 107 Excess OR b (95% CI c ) at 100 Bq/m (0.01, 0.22) Adjusted for measurement error (COV f =50%): 0.17 (0.03, 0.37) Sweden: nonsmokers [14] M+F 65.5(76% for controls) (-0.05, 1.05) Finland: south [9] M (-0.21, 0.78) Finland: nationwide [10] %M+7%F 15 (90% for controls) (-0.06, 0.31) Germany: Western [13] M+F 50(60) (-0.18, 0.17) 0.13 (-0.12, 0.46) Germany: Eastern [12] M+F (-0.03, 0.20) United Kingdom [11] M+F 99% (-0.03, 0.20) Czech Republic [16] M+F (0.02, 0.21) Spain [19] M+F (0.12, 4.48) France [15] M+F (-0.01, 0.11) for all 0.07 (0, 0.14) for complete measurements Italy [20] M+F (-0.11, 0.46) 29

48 Table 2.3 (Part I continued) China Study Enrollment Period No. of Cases No. of Controls Sex a Alive Cases Proportion (%) Mean Radon (Bq/m 3 ) Excess OR b (95% CI c ) at 100 Bq/m 3 Shenyang [17] F (-0.23, 0.19) Gansu [18] M+F 46 (96% for controls) (0.03, 0.40) Table 2.3 (Part II) Study North America New Jersey-I+II [1] Dosimetry Type and Duration Glass Detector (Y/N) g Location of Placement Average No. of Residences Subjects Mobility Within Each Residence h 1 year ATD and 4 days CC i N living areas and basement N Missouri-I [3] 1 year ATD N bedroom and kitchen 2 N Missouri-II [4] Iowa [5] 1 year ATD and 60days RRD 1 year ATD and 1 year RRD Y bedroom and kitchen N (1/2(kitchen+bedroom)) Y each level of home, current and historical bedrooms, and working areas 1 Y (current and historical bedroom, inhome work area, each level of home, outside, another building, travel) 30

49 Table 2.3 (Part II continued) Study Dosimetry Type and Duration Glass Detector (Y/N) g Connecticut [6, 135] 1 year ATD N Utah-South Idaho [6, 135] 1 year ATD N Location of Placement bedroom, another room on the lowest living level, basement, and every living level bedroom, another room on the lowest living level, basement, and every living level Average No. of Residences 4 4 Subjects Mobility Within Each Residence h Y (hours spent on various floors of the home, where they slept, whether worked outside home) Y (hours spent on various floors of the home, where they slept, whether worked outside home) Winnipeg (Canada) [2] two 6-month ATD N living areas and basement 9 N Sweden: nationwide [8] Sweden: nonsmokers [14] Finland: south [9] Finland: nationwide [10] 90 days during heating season ATD 90 days during heating season ATD 60 days during heating season ATD 1 year ATD (mail-in dosimeter installed by subjects or next of kin) N living room and bedroom 3 N N living room and bedroom N (1/2(living romm+bedroom)) N living room or bedroom 2 N N living room or bedroom N Germany: Western [13] 1 year ATD N living room and bedroom 1.1 N Germany: Eastern [12] 1 year ATD N living room and bedroom 2.4 N United Kingdom [11] 6 months ATD N living room and bedroom 3.1 N 31

50 Table 2.3 (Part II continued) Study Dosimetry Type and Duration Glass Detector (Y/N) g Czech Republic [16] 1 year ATD N Spain [19] at least 90 days (median 150 days) ATD N Location of Placement tow mostly occupied rooms one mostly occupied room (generally bedroom) Average No. of Residences Subjects Mobility Within Each Residence h 1 N France [15] 6 months ATD N living room and bedroom 2 N Italy [20] two 6-month ATD Y living room and bedroom 3 N China Shenyang [17] 1 year ATD N living room and bedroom N (higher reading used) Gansu [18] 1 year ATD N living area and sleeping area N N 32

51 Table 2.3 (Part III) Study North America Time Window Measurements Coverage in the Time Window (%) Comments New Jersey-I+II [1] 5~ First large scale study in US, radon level low Missouri-I [3] 5~ Missouri-II [4] 5~25 91 Glass detector first be used Iowa [5] 5~ Connecticut [6, 135] 5~Age25 57 Utah-South Idaho [6, 135] 5~Age25 60 Winnipeg (Canada) [2] Europe 5~30 5~15 Sweden: Stockholm [7] 5~ Inclusion criteria: subjects who had occupied their current home for at least 20 years. No imputation. No increase was detected Sweden: nationwide [8] 3~ Short period of radon measurements (3 months) Sweden: nonsmokers [14] 3~ Finland: south [9] 1950~ Only one detector per drewlling 33

52 Table 2.3 (Part III continued) Study Time Window Measurements Coverage in the Time Window (%) Finland: nationwide [10] Comments Nested case-control study within a cohort residing in the same one-family house for at least 19 years Germany: Western [13] 5~15 71 Risk was detected in radon-prone areas only Germany: Eastern [12] 5~35 72 United Kingdom [11] 5~35 85 A model was developed addressing radon exposure uncertainties Czech Republic [16] 5~34 80 Cohort study Spain [19] only current house measured France [15] 5~30 88 Italy [20] 5~34 90 China Shenyang [17] only current house >5years or the previous house measured Gansu [18] 5~30 77 Adjust for radon exposure uncertainties Note: a F=female, M=male b OR=odds ratio c CI=confidence interval 34

53 d ATD=alpha track detector e RRD=retrospective radon detector (glass-based) f COV=coefficient of variation g Y=glass detectors were used in the study, N=glass detectors were not used in the study h Y=subjects mobility within each residence was considered, N= subjects mobility within each residence was not considered i CC=charcoal container detector 35

54 36 Table 2.4 Comparison of volume traps and surface traps for 210 Pb Volume Traps Surface Traps Mechanism Radon diffusion Alpha recoil implantation Record Advantages Historical radon concentration Independent of aerosol conditions and other room parameters [137] Disadvantages The porous materials destroyed in the lab analysis [53] Much labor in lab analysis Historical deposited short-lived radon progeny The buildup of 210 Pb activity proceeds undisturbed over many years 210 Po activity is easily measured by SSNTD on the surface of chosen glass artifacts without harm to the glass Suitable for large scale field studies Dependent on room aerosol characteristics, room geometry and air movement Difficult to calibrate

55 37 Table 2.5 Correlation coefficients between implanted 210 Po activity and cumulative radon exposure obtained from literature Author Samuelsson (1992) [35] Lively and Steck (1993) [36] Mahaffey (1993) [37] Steck (2002) [59, 63, 162] Number of glass samples R > Slope (kym -1 ) Comments Complete data Excluding low quality data Radon chamber Houses Minnesota homes (20 y repeated radon measurements) Iowa homes (contemporary radon measurements) Falk (2001) [53] kym -1 was used as a constant conversion factor Samuelsson (1996) [44]

56 Figure 2.1 Factors influencing the relationship between radon exposure and lung cancer risk [21] 38

57 Figure 2.2 Human respiratory tract [21] 39

58 Figure 2.3 Percent of particles of a specific size deposited in various parts of the human respiratory tract [21] 40

59 41 Figure 2.4 Effective dose per unit exposure of the thoracic regions as function of the particle diameter [104]

60 Figure 2.5 Compartment model of the attachment, deposition and implantation of the radon progeny [114] 42

61 Figure 2.6 The bimodal distribution showing unattached and attached activity [114] 43

62 Figure 2.7 Radon decay series 44

63 Figure 2.8 Implantation by alpha recoil from radon decay products. a) Double implantation from 218 Po. b) Single implantation from 214 Po [114]. 45

64 Figure 2.9 Household glass sheets analyzed with regard to implanted 210 Po normalized to the assumed cumulative radon exposure (The dashed line is the expected 210 Po activity implanted in glass assuming decay of 210 Pb being the only loss)[114] 46

65 Figure 2.10 Different calibration curves for 210 Po in glass and integrated radon exposure in smoky versus non-smoky environments [40] 47

66 48 CHAPTER III FIELD INVESTIGATION OF THE SURFACE-DEPOSITED RADON PROGENY AS A POSSIBLE PREDICTOR OF THE AIRBORNE RADON PROGENY DOSE RATE Summary of Findings The quantitative relationships between radon gas concentration, the surfacedeposited activities of various radon progeny, the airborne radon progeny dose rate, and various residential environmental factors were investigated through actual field measurements in 38 selected Iowa houses occupied by either smokers or nonsmokers. Airborne dose rate was calculated from unattached and attached PAEC based on both Porstendörfer s effective dose conversion factor and that of James (so called Pdose and Jdose). Surface-deposited 218 Po and 214 Po were found significantly correlated to radon, unattached PAEC, and both airborne dose rates (p-values<0.0001) in nonsmoking environment, but not in smoking environment for deposited 218 Po. In multiple linear regression analysis, natural logarithm transformation was made for airborne dose rate as a dependent variable, as well as for radon and deposited 218 Po and 214 Po as predictors. In nonsmoking environment, an interaction effect was found between deposited 214 Po and an obstacle in front of the RRD in predicting dose rate (p-value= and for Pdose and Jdose respectively). After adjusting for radon and deposited radon progeny effects, the presence of either cooking, a fireplace, or a fan significantly, or marginal significantly, reduced the Pdose to 0.65 (90%CI ), 0.54 (90%CI ) and 0.66 (90%CI ), respectively. For Jdose, only the presence of fan significantly reduced the dose rate to 0.57 (90%CI ). In smoking environments, deposited 218Po was a significant negative predictor for Pdose (RR 0.68, 90%CI ) after adjusting for long-term 222Rn and environmental factors. After adjusting for the radon and radon progeny effects and other environmental factors, for every 10 increasing

67 49 cigarettes smoked in the room, a significant decrease of 0.72 (90%CI ) in the mean Pdose was found. After adjusting for the radon and radon progeny effects and other environmental factors, a significant increase of 1.71 in the mean Pdose was found for large room size relative to small room size (90%CI ), while the presence of fireplace usage was found to significantly increase the mean Pdose to 1.71 (90%CI ). Introduction In recent years, a number of large population-based epidemiologic studies have been performed in North America, Europe and China to evaluate the lung cancer risk to the general population from prolonged exposure to residential radon [1-20]. In summary, the individual studies have generally demonstrated a small, predominantly statistically non-significant, risk between residential radon gas exposure and lung cancer. While it is actually the long term exposure over past decades to the short-lived radon progeny by inhalation that dominates lung doses, for a number of practical reasons it is the contemporary radon gas that was measured in most of the existing radon and lung cancer epidemiologic studies. In the late 1980s, it was first proposed that the measurement of surface implanted 210 Po could be used as a retrospective monitor for radon exposure [33, 34]. Since then, the techniques and field methods for both accurately measuring the surface implanted 210 Po and reconstructing past radon levels in the air from the surface implanted 210 Po have been investigated by numerous scientists in Europe and the U.S. [35-61]. However, the major research challenge of this technique is reconstructing past airborne radon progeny levels from the measured surface implanted 210 Po, through effectively controlling the potentially influential environmental factors retrospectively. To improve radon progeny dose estimates, the Iowa Radon and Lung Cancer Study (IRLCS) and Missouri Radon and Lung Cancer Study-Phase II (MRLCS-II)

68 50 utilized novel retrospective radon detectors for reconstructing past radon progeny concentrations by analyzing the alpha activity deposited on and implanted in glass surfaces [36, 37, 62-64]. The surface-deposited activities of the various radon progeny reflect both the airborne radon progeny activities and atmospheric conditions. IRLCS s Retrospective Reconstruction Detector, which we call a RRD, [36, 62, 63] uses three track registration chips to simultaneously measure contemporary radon gas, contemporary surface-deposited radon progeny, and surface-implanted 210 Pb through its alpha emitting progeny 210 Po. Theoretically, contemporary airborne radon progeny concentrations can also be estimated through an empirical model using the direct measurements of the short-lived radon progeny that are deposited on surfaces and the radon gas concentration. However, realistically, significant uncertainties exist in the interpretation of the measured contemporary surface-deposited radon progeny and surface-implanted 210 Po, because the major room-specific atmospheric factors such as aerosol density and air movement can affect radon progeny deposition appreciably in houses. Due to the uncertainty in the interpretation, the IRLCS has not moved forward performing risk analyses based on the existing glass-based RRD measurements. The primary goal of the research, described below, is to accelerate the field calibration process of the RRD by characterizing the relationship between the surfacedeposited activities of various radon progeny, various residential environmental factors, and the airborne radon progeny dose rate through actual field measurements. Radon and radon progeny concentrations were measured, from the summer of 2005 to spring 2007, in 38 Iowa houses occupied by either smokers or nonsmokers. The investigation took into account several important indoor environmental factors, which were thought to have crucial influences on the radon progeny deposition process in homes. The results from our current study will aid in the final reanalysis of the lung cancer risks for the IRLCS based on radon progeny exposure estimates obtained from RRD measurements. In addition, the results of this study will be used as the basis for a large-scale pooled

69 51 analysis of the Iowa and Missouri Residential Radon Studies, both of which incorporated the use of the glass-based detectors within their study design. Methods Site selection Potential participants in Iowa were found through a local new paper in the University of Iowa Hospitals and Clinics (UIHC) and a private radon tester in Kalona, Iowa. Screening radon tests using EPERMs (Rad Elec Inc.) were performed for selected rooms in the houses of potential participants who agreed to join the study (APPENDIX A). Those houses with radon level in living area greater than 4pCi/L in screening tests and with current occupancy greater than 7 years were recruited into the final study. Homes were selected that represented a range of depositional environments. From summer 2005 to spring 2007, we screened 60 houses and identified 38 eligible houses in 13 cities or towns all within the state of Iowa for the final study. In each of the 38 houses, two frequently occupied rooms with a proper glass object were identified as study units. Active smoking was present in 18 houses. Considering the difficulties in recruiting current smokers, the eligibility criteria were slightly more relaxed for smokers. Houses of smokers with radon concentrations in the living area exceeding 2pCi/L in screening tests and with current occupancy greater than 5 years were included in the study. Description of radon and radon progeny detectors All of the novel radon or radon progeny detectors utilized in this project were fabricated and calibrated in the physics laboratory of Dr. Daniel Steck, Physics Department, St. John s University, Collegeville, Minnesota. Retrospective radon detector (RRD) The RRD utilized in this study is a track registration CR-39 chip held in a 35mm 35mm slide mounted with an 18mm 22mm opening. The energy absorbing

70 52 strips are landscape oriented. There are two sides of the detector as shown in Figure 3.1. The room side is designed to receive the alpha particles emitted from the surfacedeposited radon progeny, which is distinguished by the two strips of plastic films that cover the opening. The area covered by 3M114 film is the 214 Po sensitivity area, while the area covered by CI92 film is the 214 Po and 218 Po sensitivity area. The glass side is designed to receive the alpha particles emitted from the implanted radon progeny in glass, which is distinguished by the aluminized Mylar that cover the opening. Underneath the Mylar, the bottom two-thirds is covered by 3M2708 film that define the 210 Po sensitivity area, while the uncovered area is more sensitive to natural contaminants in glass. In field measurements, the whole slide with a CR-39 chip inside is attached to the selected glass by a tab using 3M mounting tape. Airborne radon progeny detector (ARPD) The ARPD is an active detector measuring the short-lived alpha emitting radon progeny 214 Po and 218 Po. Since the aerosol size distribution is an important factor determining the radon progeny dose estimation, the ARPD is designed to distinguish two size fractions: unattached (1 to 5 nm) and attached ( nm). As shown in Figure 3.2, the unattached radon progeny are collected by the screen, while the attached are collected by the filter. One centimeter above the screen or filter is the CR-39 chip holder, whose opening is covered by two strips of plastic films, each defining either the 214 Po sensitivity area or the 214 Po and 218 Po sensitivity area. Radon/thoron gas detector (GD) As shown in Figure 3.3, the radon/thoron gas detector contains two separate units, measuring radon and thoron gas respectively. It is a passive alpha track detector (ATD), which also utilizes the CR-39 chip as the alpha track registration media.

71 53 Radon and radon progeny measurements In each selected room in the eligible house, a proper glass object was identified according to the criteria mentioned in CHAPTER II that have been suggested in order to use glass as a retrospective monitor [47]. After the proper glass object was identified, the deposited alpha activities and the implanted activities were measured using a RRD for 90 days; and simultaneously, the radon and thoron gas concentrations were measured using a GD as indicated in Figure 3.4. Toward the end of the 90-day measurement period, for practical reasons (e.g., traveling and house owners schedules) the airborne radon progeny by species and bimodal size fraction were measured using an ARPD for 3 to 7 days; and simultaneously, the radon gas concentration was measured again using EPERMs. Collecting pertinent non-radiation data For each selected house, the information about the age of the house and years of current owner s occupancy were collected. For each selected room, the following information was either collected by questionnaire or measured by special portable devices (APPENDIX B): 1) Room type and room location within the house 2) Room size 3) Room ventilation mechanisms 4) Room cooling and heating systems 5) Room smoking frequency, other aerosol sources and aerosol sinks (e.g., cooking, candle, fireplace, humidifier, freshener, air cleaner) 6) Glass surface information (type, age, cleaning frequency, film, obstacle and air flow rates vertical/horizontal/normal to the surface) Figure 3.5 shows a typical study room where both radiation and non-radiation environmental data were collected.

72 54 Detectors processing, reading and results translating After the field measurements, the short-term radon concentration results from EPERMs were calculated. The CR-39 chips from the RRD, ARPD, and GD were disassembled in the laboratory and developed at 75 o C ±3 o C in a 6.25N NaOH solution for 6 hours and then the number of alpha tracks on track-bearing areas of each chip were read under a microscope at 100X until at least 150 tracks in three or more distinct regions were counted. The number of alpha tracks counted was then converted to tracks/mm 2 and divided by the exposure duration of each detector. Finally, the track densities were translated into radon or radon progeny activities and potential alpha energy concentrations (PAEC) with the calibration factors obtained from Dr. Daniel Steck s laboratory. The detailed calculations regarding the track-density translations of ARPD are attached in APPENDIX C. Effective lung dose estimates The deposition of radon decay products is not equal in each of the respiratory regions. The bronchial region receives the major proportion of the total lung dose in the size range of cluster mode (<5 nm), while the bronchiolar region receives the highest dose in the size range of the aerosol attached mode (>10 nm). Even in the size range from nm, where the alveolar region has relatively higher particle deposition efficiency, the effective dose contribution of this region is more than five times lower than the contribution of the bronchial and bronchiolar region. In addition, 90-95% of malignant primary lung tumors are bronchogenic. Taking into account the data regarding cancer of the human respiratory tract, Porstendörfer [79, 104] set the relative sensitivity between bronchial (WBB), bronchiolar (Wbb), and alveolar-interstitial (WAI) lung region as WBB:Wbb:WAI = 0.80:0.15:0.05. This partitioning of detriment leads to higher effective doses by the decay product clusters (unattached fraction) and lower values by the aerosol fraction than the partitioning WBB:Wbb:WAI = 0.33:0.33:0.33

73 55 recommended by ICRP 66 and revisited by James et al. [79, 104]. Patton [179] summarized the James and Porstendörfer effective dose conversion coefficients as shown in Table 3.1. In this paper, we call the effective dose calculated from the dose conversion factors recommended by Porstendörfer and James Pdose and Jdose, respectively. The Pdose and Jdose calculation formula can also be found in APPENDIX C. Data analyses Correlation analysis technique was used to estimate the correlations among the measured radon, airborne bimodal PAECs, dose rates and surface-deposited radon progeny. Multiple linear regression models were used to identify a quantitative relationship between the airborne radon progeny dose rate and the surface-deposited radon progeny as well as to estimate the effect of important indoor environmental factors. Because the radon concentrations, airborne dose rates, and surface-deposited radon progeny activities are all highly right-skewed, natural logarithm transformations were applied to all of the above variables so that parametric regression analysis could be used. Normality of each above variable was checked before and after natural logarithm transformation. Model selection based on adjusted R-square was conducted to choose the best surface-deposited radon progeny or combinations in predicting dose rate. In addition, the backward variable selection method was used to choose other potential environmental predictor variables for inclusion. Variables with p-values < 0.20 were kept in the variable selection. Analyses were carried out using SAS statistical software (SAS Institute Inc., Cary, NC, USA). Since the behavior of the radon progeny is highly related to aerosol density and smoking is the most effective indoor aerosol source, all analyses were conducted for smoking rooms and nonsmoking rooms separately. Here smoking rooms are defined if currently at least a few cigarettes are actually smoked in those rooms. Another important reason for grouping by smoking status is the artificial interaction between radon level and

74 56 smoking. By study design, smokers houses tended to have lower radon concentrations due to the less strict recruiting criteria. Results From summer 2005 to spring 2007, the complete set of radon/radon progeny and environmental factors measurements were achieved for a total of 76 rooms from the 38 Iowa houses. Among the 76 rooms, a convenience sample of 22 rooms from 11 houses was selected for repeat measurements during different seasons. The original measurements were carried out in the winter of 2005 and the repeat measurements were carried out in the summer of 2006 for these 22 rooms. Information on all environmental factors that were included in the original measurement period was collected during the repeat measurement period as well. Because the primary research interest in this chapter is to study the physical behavior of radon progeny within each room through simultaneous measurements of both airborne and surface-deposited radon progeny, all 98 sets of data were included for the final analysis. Unfortunately, 26 of the 98 data sets were excluded from the analysis involving surface-deposited radon progeny, because the 218 Po surface activities given by the RRDs were negative for those rooms. Descriptive statistics The descriptive statistics for major physical parameters measured are summarized in Table 3.2 by smoking status. The short-term 222 Rn refers to the radon concentrations measured simultaneously with the ARPD using EPERM, while the long-term 222 Rn refers to the radon concentrations measured simultaneously with RRD using GD. As expected, the means for both the short-term 222 Rn and long-term 222 Rn from smoking rooms are much lower than those from nonsmoking rooms due to the different recruiting criteria for smokers and nonsmokers. The equilibrium factor F and unattached fraction fp, which should have no relation with radon concentration, were negatively related as we expected (Figure 3.6). In addition, smoking was positively related to F and negatively related to fp.

75 57 The descriptive statistics for major environmental factors are summarized in Table 3.3. The median residency in current homes is 12.5 years and the median age of the houses is 43 years. Correlation analysis Table 3.4 presents the Spearman correlation coefficients between airborne and deposited measurements by smoking status. Nonsmoking rooms In nonsmoking rooms, both deposited 218 Po and deposited 214 Po activities have very good correlation with short-term 222 Rn, long-term 222 Rn, airborne unattached PAEC, Pdose, and Jdose (p-value<0.0001). Only the airborne attached PAEC was not significantly correlated with deposited radon progeny. The slow deposition velocity for large size particles could well explain this phenomenon. For airborne dose rate, the correlation between Pdose and the deposited radon progeny is slightly better than that between Jdose and deposited radon progeny, because Pdose puts higher weight on unattached fraction of PAEC. Smoking rooms In smoking rooms, which represent a low deposition environment, none of the Spearman correlations between deposited 218 Po and airborne parameters is significant at the 0.05 level. Alternatively, the correlations between deposited 214 Po and airborne parameters remain significant with exception of the attached PAEC. Regression analyses Normality check In order to fulfill the assumption of multiple linear regression method, the dependent variable (in our study, Pdose and Jdose) has to be normally distributed. A

76 58 Shapiro-Wilk test was used to check the normality. After the natural logarithm transformation, tests for both lnpdose and lnjdose failed to reject the normality none hypothesis at the 0.05 level of significance (lnpdose: p-value=0.3366, lnjdose: p- value=0.0653), which indicated that lnpdose and lnjdose could be treated as normally distributed and thus eligible to be a dependent variable in the multiple linear regression (Figure 3.7 and Figure 3.8). Although there is no normality requirement for explanatory variables, normality or near normality of explanatory variables will improve the accuracy of the regression parameters. Normality was also tested for ln long-term 222 Rn, ln deposited 218 Po and ln deposited 214 Po with p-values , and , respectively. In following regression analyses, long-term 222 Rn, deposited 218 Po and deposited 214 Po are all in natural log-scale. Selection of deposited radon progeny Regression analyses were divided into two steps. First, only long-term 222 Rn, deposited 218 Po, and deposited 214 Po were considered as potential predictors of airborne dose rate (Pdose and Jdose). The selection results based on the adjusted R-square are presented in Table 3.5. When the backward variable selection method was used, it produced the same selection results as the adjusted R-square, see Table 3.5. The adjusted R-square for short-term 222 Rn was also used for comparison with long-term 222 Rn. The higher correlation between dose rate and short-term 222 Rn as compared to long-term 222 Rn likely reflects the fact that the dose rate and short-term 222 Rn were measured simultaneously for 3-7 days during the end of the 90-day long-term measurements. For nonsmoking rooms, the combination of long-term 222 Rn and deposited 214 Po performs the best for predicting both Pdose and Jdose. For smoking rooms, the combination of longterm 222 Rn, deposited 218 Po, and deposited 214 Po performs best in predicting Pdose, while the combination of long-term 222 Rn and deposited 218 Po is best choice for predicting Jdose.

77 59 Selection of environmental factors The backward variable selection method was used to choose environmental factors that have potential influence on the relationship between airborne dose rate and deposited radon progeny. The a priori selected environmental factors for this analysis were room size (large or small), presence/absence of cooking, presence/absence of fireplace usage, presence/absence of candle burning, presence/absence of air cleaner usage, presence/absence of window opening, presence/absence of central air system, presence/absence of fan usage, presence/absence of obstacles in front of the RRD, and the presence of dust film on the glass surface on a scale of Four separate models were built for Pdose and Jdose and by different smoking status. When selecting the environmental factors, the best combinations of radon and deposited radon progeny decided in the first step were forced to be included in each regression model. The results of the model selection for Pdose are shown in Table 3.6 and the results for Jdose are presented in Table 3.7. For both Pdose and Jdose, under both smoking and nonsmoking conditions, the long-term 222 Rn was a significantly positive predictor to dose rate at 0.10 significance level after adjusting for all other covariates in the model (Table 3.6 and Table 3.7). In nonsmoking environments, for both Pdose and Jdose, significant interaction effect was detected at 0.10 level between deposited 214 Po and obstacle in front of RRD in predicting dose rate (p-value= and for Pdose and Jdose respectively). When there was no obstacle in front of RRD, the ln deposited 214 Po was positively related to lnpdose; while when an obstacle existed, no observable relationship was found between deposited 214 Po and lnpdose (Figure 3.9). The same findings were noted for Jdose. Presence of cooking, a fireplace, or a fan reduced, either significantly or with marginal significance, the Pdose after adjusting for other predictors in the model including the interaction effect between deposited 214 Po and an obstacle. For example, the presence of cooking in the room significantly reduced the Pdose to 0.65 (90%CI 0.42-

78 ). For Jdose, only the presence of fan significantly reduced the dose rate to 0.57 (90%CI ). In smoking environments, deposited 218 Po was a significant negative predictor for Pdose (RR 0.68, 90%CI ) after adjusting for long-term 222 Rn and environmental factors. After adjusting for the radon and radon progeny effects and other environmental factors, for every 10 increasing cigarettes smoked in the room, a significant decrease of 0.72 (90%CI ) in the mean Pdose was found. After adjusting for the radon and radon progeny effects and other environmental factors, a significant increase of 1.71 in the mean Pdose was found for large room size relative to small room size (90%CI ). After adjusting for the radon and radon progeny effects and other environmental factors, the presence of fireplace usage was found to significantly increase the mean Pdose to 1.71 (90%CI ). Very similar results were also found for Jdose. Discussions and Conclusions From the multiple regression analyses, we found that the surface-deposited radon progeny were useful in predicting airborne dose rate even after adjusting for the radon gas effect in both nonsmoking and smoking environments. In the nonsmoking environment, Pdose exhibited greater sensitivity to additional aerosol sources like cooking and fireplace as compared to Jdose, because Pdose was weighted more toward the unattached fraction of PAEC that would be reduced by the presence of additional aerosol sources. In addition, the usage of a fan was found to effectively reduce both Pdose and Jdose by accelerating radon progeny deposition and exhaust in the nonsmoking environment. In the smoking environment, as the number of cigarettes smoked in the room increased, both Pdose and Jdose decreased. The underlying fact could be that when the aerosol density increased, more radon progeny were attached to the aerosols, which have lower effective dose conversion factor. After adjusting the smoking effect, fireplace usage became a positive predictor for both Pdose and Jdose.

79 61 The fact that larger room sizes have smaller surface-to-volume ratios and thus lower deposition rates for both unattached and attached radon progeny likely contributes to the positive effect of room size on both Pdose and Jdose. The interaction effect between obstacle objects in front of the glass and the deposited radon progeny in predicting airborne dose rate indicates that in the future epidemiologic studies involving radon progeny exposure assessment using RRD, the glass with obstacle objects in front of it should be avoided as the RRD placement site. Because the study houses were selected based on convenience rather than a random sample of Iowa houses, a potential limitation of the study is the generalizability of the study results to a wide cross-section of homes. It is unknown how well the selected homes represent the general housing stock in Iowa. Nevertheless, the study was focused on objective measurements of radon and radon progeny as well as related environmental factors including fan usage, heating type, etc. The heating/ventilating/air conditioner (HVAC) system represents the major air ventilation style in the house. In current field study, 81.6% of the houses have forced air heating/cooling (central AC) systems. While in the original IRLCS, 80.4% of the controls houses have forced air heating/cooling systems [5]. From the measurement perspective, a potential bias might be introduced by the fact that the ARPD only measures a small portion of the entire measurement period of the RRD. Because long-term active airborne radon progeny concentrations can not be measured inexpensively and accurately with current technology, we can only perform short-term active airborne radon progeny measurements. In addition, in order to minimize the total visiting times to the study houses, the ARPD measurements usually took place toward the end of the 90 days, thus all field measurements, including RRD could be finished at the same time. The last 3 to 7 days of the 90 days may not represent the typical patterns of weather and related heating/cooling system usage in the house. Differences in the depositional environment and radon concentrations over the 90 days as

80 62 compared to the final 3 to 7 days may introduce further error when matching the results from the RRD and ARPD. Nevertheless, measurements of the actual airborne radon progeny activity represent a significant advance over previous epidemiologic studies or field investigations. Even the results from short-term ARPD measurements will add invaluable information to our understanding of radon progeny behavior in the actual field setting. Because the majority of information on the environmental factors (e.g. smoking frequency) that could influence the radon and radon progeny measurements was obtained by questionnaire, both recall bias and information bias may have occurred. In addition, some environmental factors like aerosol conditions and ventilation usage were constantly changing from time to time during the 90-days study period, which made estimating the average over the time period more prone to error. To improve the reliability of the collected information, the same questions on smoking and other important environmental factors were asked at the start of the study and then asked again at the end of the 90 days. And it turned out that the reliability for the collected environmental information was quite good between the first and second time points. Detailed investigations examining the relationship between airborne radon progeny concentrations and the measured surface-deposited radon progeny concentrations, considering multiple influential environmental factors such as ventilation and different sources of indoor aerosols, are lacking in the field setting. Therefore, our field investigation of the performance of the novel glass-based retrospective radon progeny reconstruction detector (RRD), in conjunction with simultaneously using several other novel alpha track detectors like the airborne radon progeny detector (ARPD) and radon/thoron gas detector, will benefit both scientific knowledge and physical the technology regarding the radon dosimetry thus leading to better estimation of lung cancer risk posed by prolonged residential radon progeny exposure.

81 63 Table 3.1 Calculated effective dose conversion factors (DCF) [179] Effective DCFs (msvyr -1 WLM -1 ) nm nm James Porstendörfer

82 64 Table 3.2 Descriptive statistics for measured radon, airborne PAECs, dose rates, F, fp and surface deposited radon progeny by smoking status Variables Environment (n) Range Median Mean (SD) Unit Short-term 222 Rn Unattached PAEC Attached PAEC Pdose Jdose F Fp Long-term 222 Rn Deposited 218 Po Deposited 214 Po Nonsmoking (62) (218) Smoking (36) (111) Total (98) (200) Nonsmoking (62) (3.0) Smoking (36) (1.0) Total (98) (2.7) Nonsmoking (62) (16.6) Smoking (36) (13.4) Total (98) (15.4) Nonsmoking (62) (31.4) Smoking (36) (11.6) Total (98) (28.1) Nonsmoking (62) (19.7) Smoking (36) (10.1) Total (98) (17.5) Nonsmoking (62) (0.16) Smoking (36) (0.18) Total (98) (0.17) Nonsmoking (62) (0.26) Smoking (36) (0.18) Total (98) (0.25) Nonsmoking (62) (172) Smoking (36) (105) Total (98) (157) Nonsmoking (48) (48.3) Smoking (24) (18.7) Total (72) (41.5) Nonsmoking (48) (22.5) Smoking (24) (10.6) Total (72) (20.2) Bqm -3 mwl mwl msvyr -1 msvyr -1 Bqm -3 Bqm -2 Bqm -2

83 65 Table 3.3 Descriptive statistics for environmental categorical variables Variable Levels N Percents Season spring summer winter Room type basement 4 5 bedroom dinning room 6 8 entertainment room 3 4 family room hallway 1 1 kitchen 5 7 living room work room 5 7 Room size small large Room level basement first floor second floor 4 5 Smoking no yes Cooking no yes Candle no yes Fireplace no yes Humidifier no yes 4 5 Air cleaner no yes 2 3

84 66 Table 3.3 (Continued) Variable Levels N Percents Open window no yes HVAC none forced air window AC 1 1 Fan usage none ceiling fan floor fan 2 3 other fan 2 3 Note: HVAC=heating, ventilation and air conditioner

85 67 Table 3.4 Spearman correlation coefficients and p-values under zero-correlation none hypothesis between airborne and deposited measurements by smoking status Environment Nonsmoking (48) Smoking (24) Airborne measurements Short-term 222 Rn 0.58 <.0001 Long-term 222 Rn 0.62 <.0001 Unattached PAEC 0.61 <.0001 Attached PAEC Pdose 0.59 <.0001 Jdose 0.53 <.0001 Deposited radon progeny 218 Po 0.73 < < < < Deposited 218 Po 0.58 <.0001 Short-term 222 Rn Long-term 222 Rn Unattached PAEC Attached PAEC Pdose Jdose Deposited 218 Po Po

86 68 Table 3.5 Adjusted R-square from the natural log-scale a multiple regression results for deposited radon progeny in comparison to radon Environment Dose being predicted Shortterm Rn b Longterm Rn c Long-term Rn and deposited 214 Po c Long-term 222 Rn and deposited 218 Po c Long-term 222 Rn and deposited 214 Po, 218 Po c Nonsmoking Pdose d Jdose Smoking Pdose Jdose Note: a All dependent variables and explanatory variables in the regression analysis are in natural log-scale. b The whole data set (Total=98, Nonsmoking=62, Smoking=36) was used. c The valid deposited radon progeny data set (Total=72, Nonsmoking=48, Smoking=24) was used. d The bold italics values indicate the adjusted R-square for the best model selected.

87 69 Table 3.6 Estimated relative mean change in Pdose for selected predictors using multiple regression analysis by smoking status Environment Predictor (N) Relative mean Pdose (90% CI) a Nonsmoking Obstacle=0 Obstacle=1 Smoking Ln Long-term 222 Rn (unit 1 increase) (48) 1.57 ( ) Ln Deposited 214 Po (unit 1 increase) (41) 1.62 ( ) Ln Deposited 214 Po (7) 0.12 ( ) (unit 1 increase) Cooking (48) Present (5) 0.65 ( ) not present (43) 1.00 b Fireplace (48) Present (2) 0.54 ( ) not present (46) 1.00 b Fan (48) Present (8) 0.66 ( ) not present (40) 1.00 b Ln Long-term 222 Rn (24) 2.28 ( ) (unit 1 increase) Ln Deposited 218 Po (24) 0.68 ( ) (unit 1 increase) Number of cigarette (24) 0.72 ( ) (10 increase) Size (24) Large (16) 1.47 ( ) Small (8) 1.00 b Fireplace (24) Present (5) 1.71 ( ) not present (19) 1.00 b a The relative mean change in Pdose is after adjusting for all other covariates in each smoking status in the table and the regression equations by smoking status are attached in APPENDIX D. b Reference category

88 70 Table 3.7 Estimated relative mean change in Jdose for selected predictors using multiple regression analysis by smoking status Environment Predictor (N) Relative mean Jdose (90% CI) a Nonsmoking Obstacle=0 Obstacle=1 Smoking Ln Long-term 222 Rn (unit 1 increase) (48) 1.56 ( ) Ln Deposited 214 Po (unit 1 increase) (41) 1.51 ( ) Ln Deposited 214 Po (7) 0.16 ( ) (unit 1 increase) Fan (48) Present (8) 0.57 ( ) not present (40) 1.00 b Ln Long-term 222 Rn (24) 2.42 ( ) (unit 1 increase) Ln Deposited 218 Po (24) 0.69 ( ) (unit 1 increase) Number of cigarette (24) 0.74 ( ) (10 increase) Size (24) Large (16) 1.55 ( ) Small (8) 1.00 b Fireplace (24) Present (5) 1.75 ( ) not present (19) 1.00 b a The relative mean change in Jdose is after adjusting for all other covariates in each smoking status in the table and the regression equations by smoking status are attached in APPENDIX D. b Reference category

89 Figure 3.1 The schematic drawing of RRD 71

90 Figure 3.2 The schematic drawing of ARPD 72

91 Figure 3.3 Photo of the radon/thoron gas detector 73

92 Figure 3.4 The glass-based RRD and radon/thoron gas detector are simultaneously exposed for 90 days in a selected room 74

93 Figure 3.5 A typical study room with radon/radon progeny detectors and relevant environmental factors distribution 75

94 Figure 3.6 The relationship between equilibrium fraction F and unattached fraction fp 76

95 Figure 3.7 Histograms of Pdose before and after log transformation 77

96 Figure 3.8 Histograms of Jdose before and after log transformation 78

97 Figure 3.9 The interaction effect between obstacle and deposited 214 Po in predicting airborne Pdose 79

98 80 CHAPTER IV FIELD EVALUATION OF TEMPORAL AND SPATIAL VARIATIONS OF THE AIRBORNE RADON PROGENY DOSE RATE AND DEPOSITED RADON PROGENY Summary of Findings The temporal variations, as well as the within-house and between-houses variations of radon concentrations, of the airborne radon progeny dose rate and deposited radon progeny activities were evaluated through actual field measurements and subgroup repeated measurements during different time periods. Environmental factors that have potential influences on the variations of airborne dose rate and deposited radon progeny were also studied. The total variation, between-house variation, within-house variation, and within room temporal variation for Pdose were 103%, 74%, 58%, and 60%, respectively. The variations for Jdose were 100%, 66%, 61%, and 46%, respectively. For deposited 214 Po, the variations were 79%, 57%, 42%, and 48%, respectively. Smoking was found to be the most important environmental factor affecting both temporal and spatial variations in airborne radon progeny dose rate. Introduction Epidemiologic studies that investigate the relationship between prolonged radon and lung cancer require accurate estimates of the long-term average concentrations of radon progeny in dwellings. The usage of contemporary radon gas concentrations as a surrogate for radon-related dose has the potential to introduce significant uncertainty in dose assessment [63]. The error induced by estimating effective dose from measured radon gas concentrations is much higher than that if the effective dose is estimated from measured radon progeny values [144]. Better dose estimation could be achieved by measuring radon as well as radon progeny concentrations under the same circumstances. However, data describing the simultaneous measurements of radon and its progeny are

99 81 sparse. Accurate dose assessment requires not only knowledge of the radon and progeny concentrations, but also information regarding the impact of human activities on the overall exposure. Radon and its progeny are subject to both significant spatial [145] and temporal variations [140, , 151, 152, ] in domestic environments, which poses additional obstacles for reconstructing an individual's long-term radon-related exposure. The primary aims of this study were to assess the temporal variation, as well as the within-house and between-house variations of the radon concentrations, the airborne radon progeny dose rate, and deposited radon progeny activities. An additional aim of the study was to examine the environmental factors that have potential influences on the above variations. Methods Study design Between the summer of 2005 and spring 2007, 38 volunteers provided access to their homes so that we could assess radon progeny behavior within the homes. The houses were located in Southeast Iowa and occupied by either active smokers or lifetime nonsmokers. The houses were selected within a larger population of homes on the basis of screening radon tests indicating they contained elevated radon concentrations in the living area. The majority of the homes were occupied by the current resident for at least 5 years. In each of the 38 houses, two frequently occupied rooms with an appropriate glass object were identified. The study involved placing a novel glass-based retrospective radon detector (RRD) on an appropriate glass object for a 3-month period and placing an airborne radon progeny detector (ARPD) for up to a week in each identified eligible room in the house. Contemporary radon gas concentrations were measured simultaneously with both the RRD and the ARPD during their individual testing periods. Domestic environmental parameters, such as indoor aerosols and

100 82 ventilation conditions, were investigated through face-to-face interview. Detailed information regarding the field study of radon progeny behavior can be found in Chapter III. In order to evaluate the temporal variation of radon and radon progeny, a sample of eleven houses were selected from the 38 houses in the main study that was described in Chapter III. Most of the 11 homes selected for this phase of the study were located in Kalona, Iowa for convenience of sampling. Current smokers were present in 3 of the 11 houses. The original tests of these 11 houses mainly took place in winter 2005 or spring All the radon and radon progeny measurements and the concurrent indoor environmental condition assessments were repeated in the summer and fall of 2006 in the selected 11 houses according to the same protocols of the original investigation. Data analysis The linear mixed model was used to assess temporal, as well as within-house and between-houses variances in the radon concentrations, airborne PAECs, radon progeny dose rate, and deposited radon progeny. The Maximum Likelihood method was used to estimate the variance components. The house and room were treated as random factors and the level of room within house and other environmental variables were treated as fixed factors in analysis. A new concept of the coefficient of variation (COV) was defined as a measure of variation: COV = 100 σ / X (4.1) Where σ is the square root of the estimated variance component from the linear mixed model. Simple linear regression was used to evaluate the correlation between repeated measurements of radon concentrations, airborne PAECs, radon progeny dose rate, and deposited radon progeny from different time periods.

101 83 Univariate analysis, using a linear mixed model, was used to identify important environmental factors affecting either spatial or temporal variations in airborne dose rate and deposited radon progeny. Analyses were carried out using SAS statistical software (SAS Institute Inc., Cary, NC, USA). Results and Discussions Among all 98 sets of complete measurements, including both 76 original sets of measurements and 22 repeated sets of measurements, twenty-six sets of measurements had to be excluded from the analysis involving surface-deposited radon progeny, because their deposited 218 Po concentrations were undetected by the RRD. Spatial variations To assess the spatial variations, only the 76 original sets of measurements were included. To assess the variations of deposited radon progeny, only 61 sets of measurements were considered valid for deposited 218 Po. The spatial variations for shortterm and long-term 222 Rn, airborne PAECs, radon progeny dose rate, and deposited radon progeny by smoking status are presented in Table 4.1. From the total variations that include not only spatial variation among different houses, but also all random measurement error variations introduced by environmental factors and variations from unknown reasons, we can see the variation in long-term radon concentrations was slightly smaller than that in short-term radon concentrations (COV 64% versus 79%). However, the airborne PAEC attached and unattached, airborne dose rate (Pdose and Jdose) all have higher total variations (COV>100%) than the radon gas concentrations. This phenomenon reflects the fact that airborne radon progeny concentrations are more sensitive to the changes in major room-specific atmospheric factors such as aerosol density, air movement, and human activities than radon gas concentrations. The total variation for deposited 214 Po was lower as compared to deposited 218 Po (COV 79% versus 117%) and lower than the airborne PAECs as well, but similar to the total radon variation.

102 84 While the house-to-house variations reflect the spatial variations for the field measurements, the within-house variations reflect not only room-to-room variations, but also within-room variations and detector measurement errors. As can be seen from Table 4.1, generally large within-house variations exist for most variables. Three variables, unattached PAEC, Pdose and Jdose, have larger total variations than the subgroup variations in both smoking and nonsmoking categories, which may indicate that smoking, as a major source of indoor aerosol, could explain a certain fraction of total variations in airborne unattached PAEC and dose rate. It should also be noted that the within-house variation for radon, attached PAEC, and dose rate were lower in smoking rooms than in nonsmoking rooms, which could imply a more stable atmosphere within smoking houses where the majority of radon progeny may attach to aerosols. Temporal variations To assess the temporal variations, the 22 rooms with repeated measurements (44 measurements) from different time periods were included in the analysis. For examination of the variations of deposited radon progeny, 33 measurements were considered valid for deposited 218 Po. The temporal variations for short-term and longterm 222 Rn, airborne PAECs, radon progeny dose rate, and deposited radon progeny by smoking status are presented in Table 4.2. Somewhat counter intuitively, the long-term radon concentrations yielded larger temporal variation than the short-term radon concentrations as can also be seen from Figure 4.1 (COV 55% versus 37%, R versus 0.49). Further examination of the data found two outlier rooms, both belonging to the same house. The original long-term radon concentrations measured in winter for these two rooms were 233 and 359 Bqm -3. ; while the repeated measurements in summer provided the results 37 and 19 Bqm -3, respectively; resulting in COVs of 103% and 128%, respectively. The likely reason for the large differences between the two long-term radon

103 85 measurements for each room is that in winter, the whole house was closed tightly 24 hours per day; while in summer, all windows of the house were open 24 hours per day. The special case illustrates that even a 90-day long-term domestic radon gas measurement is subject to large variations due to the changes in human behavior. For each room in each test season, one short-term radon measurement was performed at the end of the long-term measurement period, which provides another measurement from a different time period. The correlation (R 2 =0.53) between short-term radon measurements and long-term radon measurements is presented in Figure 4.1. The correlations between repeated PAEC measurements were not as strong as that of radon (Table 4.2 and Figure 4.2). Attached PAEC has slightly better correlation than unattached PAEC (COV 61% versus 67%, R 2 =0.3 versus 0.2), which could explain the better correlation between repeated Jdose as compared to Pdose (COV 46% versus 60%, R 2 =0.47 vs 0.27) as shown in Table 4.2 and Figure 4.3. Similar to the spatial variation findings, deposited 214 Po also displayed less temporal variation as compared to deposited 218 Po and airborne PAECs. A correlation plot was not performed for deposited 218 Po due to the large fraction of invalid values. For all variables listed in Table 4.2, the temporal variations noted for the rooms where smoking occurred were less than those of nonsmoking rooms. Environmental factors affecting the between-room and within-room variations As shown in Table 4.3 and Table 4.4, adding smoking to the model reduced both room-to-room variance and residual variance, which includes temporal variation and remained variation, in Pdose and Jdose. The variation observed between-rooms is likely attributable to a large extent to the inherent differences between smoking rooms and nonsmoking rooms in regard to aerosol condition. A large contribution of the withinroom variation in the current analysis was attributable to a room that was smoked in

104 86 during the initial testing and that became a room that was no longer smoked in the second measurement period due to the death of the individual who smoked. Fan and fireplace usage could also explain some of the temporal (or within-room) variation in both Pdose and Jdose, but could not reduce the between-room variance in the model, because fan and fireplace usage has very obvious seasonal patterns. However, room level and room type could only reduce between-room variance since they are fixed. For deposited 214 Po (Table 4.5), smoking, fireplace usage, and room level behaved the same way as for Pdose and Jdose. Opening windows and candle usage could reduce both between-room variance and temporal variance in the model because they are relevant to indoor room aerosol changes and have seasonal patterns. For example, householders tend to open window more frequently in summer and burn candles more during Fall and Winter holidays. The implications from the environmental factors selected by the current analysis should be cautiously interpreted due to the limited statistical power caused by the small sample size (22 rooms, 2 measurements per room).

105 87 Table 4.1 The spatial coefficient of variation (COV) for short-term and long-term 222 Rn, airborne PAECs, dose rate, and deposited radon progeny by smoking status Variables Short-term 222 Rn Long-term 222 Rn Unattached PAEC Attached PAEC Pdose Jdose Deposited 218 Po Deposited 214 Po Environment (n) Mean a Total % House to house % Within house b % Nonsmoking (44) Smoking (32) Total (76) Nonsmoking (44) Smoking (32) Total (76) Nonsmoking (44) Smoking (32) Total (76) Nonsmoking (44) Smoking (32) Total (76) Nonsmoking (44) Smoking (32) Total (76) Nonsmoking (44) Smoking (32) Total (76) Nonsmoking (40) Smoking (21) Total (61) Nonsmoking (40) Smoking (21) Total (61) Note: a The units for the means of the radon concentration, PAEC, dose rate and deposited radon progeny are Bqm -3, mwl, msvyr -1 and Bqm -2 respectively. b The results for within-house COV are after adjusting for the fixed effect of a room s level within a house.

106 88 Table 4.2 The temporal coefficient of variation (COV) for short-term and long-term 222 Rn, airborne PAECs, dose rate, and deposited radon progeny by smoking status Variables Environment (n) Mean a Temporal variation between two repeated measurements % Short-term 222 Rn Long-term 222 Rn Unattached PAEC Attached PAEC Pdose Jdose Deposited 218 Po Deposited 214 Po Nonsmoking (35) Smoking (9) Total (44) Nonsmoking (35) Smoking (9) Total (44) Nonsmoking (35) Smoking (9) Total (44) Nonsmoking (35) Smoking (9) Total (44) Nonsmoking (35) Smoking (9) Total (44) Nonsmoking (35) Smoking (9) Total (44) Nonsmoking (25) Smoking (8) Total (33) Nonsmoking (25) Smoking (8) Total (33) Note: a The units for the means of the radon concentration, PAEC, dose rate, and deposited radon progeny are Bqm -3, mwl, msvyr -1 and Bqm -3, respectively.

107 89 Table 4.3 Univariate analysis of environmental factors as fixed factors affecting variation in Pdose Variable P-value for F test of fixed effect a Room to room variance smoke fan fireplace level humidifier room type Note: a Fixed effects with p-value<0.2 were included in the table. Residual variance

108 90 Table 4.4 Univariate analysis of environmental factors as fixed factors affecting variation in Jdose Variable P-value for F test of fixed effect a Room to room variance smoke fan fireplace level humidifier room type Note: a Fixed effects with p-value<0.2 were included in the table. Residual variance

109 91 Table 4.5 Univariate analysis of environmental factors as fixed factors affecting variation in deposited 214 Po Variable P-value for F test of fixed effect a Room to room variance smoke fireplace openwindow candle level Note: a Fixed effects with p-value<0.2 were included in the table. Residual variance

110 Figure 4.1 Correlations between different radon measurements 92

111 Figure 4.2 Correlations between repeated airborne PAEC measurements 93

112 Figure 4.3 Correlations between repeated airborne dose rate measurements and surfacedeposited 214 Po measurements 94

113 95 CHAPTER V ROOM MODEL BASED MONTE-CARLO SIMULATION STUDY OF THE RELATIONSHIP BETWEEN THE AIRBORNE DOSE RATE AND THE SURFACE-DEPOSITED RADON PROGENY Summary of Findings The quantitative relationships between radon gas concentration, the surfacedeposited activities of various radon progeny, the airborne radon progeny dose rate, and various residential environmental factors were investigated through Monte-Carlo simulation study based on extended Jacobi room model. Surface-deposited 218 Po and 214 Po were found significantly correlated to radon, PAECs, and airborne dose rate (pvalues<0.0001) in both nonsmoking and smoking environments, except that in nonsmoking environments the deposited radon progeny were not highly correlated to attached PAEC. In multiple linear regression analysis, natural logarithm transformation was made for airborne dose rate as a dependent variable, as well as for radon and deposited 218 Po and 214 Po as predictors. In nonsmoking environments, deposited 214 Po was found a significant positive predictor for Pdose (RR 1.46, 95%CI ), while deposited 218 Po was found a negative predictor Jdose (RR 0.90, 95%CI ) after adjusting for the effect of radon. In smoking environments, deposited 218 Po was found a significant positive predictor for Pdose (RR 1.10, 95%CI ), while a significant negative predictor for Jdose (RR 0.90, 95%CI ) after adjusting for radon and room size. After adjusting for radon and deposited 218 Po, a significant increase of 1.14 (95%CI ) in the mean Pdose and 1.13 (95%CI ) in the mean Jdose was found for large room sizes relative to small room sizes. Introduction The concentrations of radon and its decay products are influenced by the complex interaction of a number of processes, the most important of which are radioactive decay

114 96 (especially α-decay), ventilation, attachment to aerosols, and deposition on surfaces [65, 114]. Radon progeny can be divided into four different groups depending on their various states: attached, unattached, deposited, and implanted. The transition and interplay between these compartments can be described by the mathematical room model commencing with that of Jocobi [115], extended by Porstendörfer et al., and others [ ]. One of the useful applications of the room model is to study the behavior of radon progeny and to identify the key parameters influencing their behavior. In this work, a room model based on those of Jocobi [115] and Porstendörfer et al. [116] modified by Steck [192], incorporating the new dose conversion factors from James et al. [79] and Porstendörfer [104] and the new specific deposition velocity for unattached 214 Pb from Schmidt and Hamel [163], was developed to study the theoretical relationship between the airborne dose rate and the surface-deposited activities of radon progeny. The Monte-Carlo approach was used to produce the probability distributions of the model outputs, based on the probability distribution assumptions of all input parameters. The results from the Monte-Carlo simulation study were then compared to the results from the actual field measurements of 76 rooms in 38 Iowa houses that are described in CHAPTER III. Methods Tools and assumptions Steck [193] has developed a Monte-Carlo (MC) platform for simulating radon and lung cancer epidemiologic studies, which included a complex radon and radon progeny dosimetry method and also provided estimators of several radon exposure surrogates that were widely used by previous epidemiologic studies. The room model used in current work is adopted from Steck s work, which is capable of generating the radon progeny in unattached, attached deposited, and implanted states, as well as airborne dose rate and other measurable variables for a set of houses characterized by the radon gas distribution

115 97 and smoking prevalence obtained from Iowa Radon Lung Cancer Study (IRLCS) [59, 63, 159, 161, 162, 192]. The main assumptions for this model are the uniform-mixing or well-stirred hypothesis and the steady-state hypothesis as described by Walsh and McLaughlin [56]. The convenient formulation of Knutson [118] is used and is in terms of an activity balance. The implantation model and calculation of the implanted 210 Po activity are based on Cornelis model [158, 194] modified by Steck. The probability distribution assumptions of all input parameters are generated from either population-specific distributions of IRLCS or distributions obtained in a literature review. Crystal Ball 7.3 (Oracle, 7700 Technology Way, Denver, CO, USA) embedded within EXCEL 2003 is the basic tool for MC model generation and sensitivity analyses. Software SAS (SAS Institute Inc., Cary, NC, USA) is used for generating descriptive statistics and applying multiple linear regression analysis for the simulated data. The model algorithm is written so that it can be used (1) to estimate the airborne and surfacedeposited activities from the various radon progeny due to a known radon gas concentration; and (2) to estimate the influences of room parameters on the model outputs as shown schematically in Figure 5.1. Room parameters probability distributions The steady-state room model relies on four semi-empirical parameters; the surface-to-volume ratio, deposition rate, aerosol attachment rate, and ventilation rate [161]. Surface-to-volume ratio of a room A bimodal normal distribution was used with mean 3.5 and standard deviation 0.3 m -1 for large rooms and mean 4.5 and standard deviation 0.4 m -1 for small rooms fitted from the measured surface-to-volume ratio distribution in IRLCS houses. The ratio in IRLCS was calculated from measured room dimensions and the fraction of the floor

116 98 covered by furniture [161]. These values are somewhat higher than the typical values of 2-5 m -1 recommended by Knutson [118], but more in line with more recent recommendations from Porstendörfer [66]. Deposition velocities A large range of values for deposition velocities of both attached and unattached progeny are reported in the literature [114]. The deposition velocity for unattached 218 Po is assumed to be lognormally distributed with mean 5 and standard deviation 3 mh -1 in the current model. The deposition velocity for unattached 214 Pb and 214 Bi/ 214 Po is assumed to be 1/4 of that for unattached 218 Po according to the work of Schmidt and Hamel [163]. The deposition velocities for attached radon progeny are assumed to be 1/100 of those for their unattached counterparts according to Knutson s suggestion [118]. The deposition rate λ d for each radon progeny in each state used in the model is derived from its own deposition velocity V d and the common surface to volume ratio of the room S/ V using the following equation: d d ( S/ V) λ = V (5.1) Attachment rates The attachment rate λ a of a nuclide (or unattached radon progeny) to aerosol particles is dependent on both aerosol particle concentration Z and attachment rate coefficient β, which is a function of the particle size distribution, as shown in the following equation: λ = β Z (5.2) a Aerosols play an important role in the behavior of radon progeny. The indoor aerosol condition is a result of a complex interplay among many sources and sinks and is changing from time to time. Cooking and smoking are prominent sources of indoor aerosols, but there are many other sources as well, like burning candles and using forced

117 99 air furnaces. Because cooking and heating are common family practices, while smoking is a key factor making the indoor aerosol conditions very different between smokers and nonsmokers rooms, Steck applied the information achieved from the literature and assumed the attachment rate to be lognormally distributed with mean 91 and standard deviation 204 h -1 for nonsmoky rooms and mean 251 and standard deviation 107 h -1 for smoky rooms. Ventilation rate The lognormal distribution with mean 0.4 and standard deviation 0.2 h -1 for ventilation rate is used in this work, which is adopted from a previous Minnesota radon survey. All the above model parameter distributions and the population distributions for other input parameters generated from IRLCS are summarized in Table 5.1. The model constants and dose conversion factors are summarized in Table 5.2. The radon gas concentration distribution, fraction of large rooms, and fraction of smoking rooms are set according to the results from the field study in CHAPTER III. Therefore, the quantitative relationship between the airborne radon progeny dose rate and the alpha activity from surface-deposited radon progeny established from the field measurements could be compared to the relationship derived from the MC model simulations, since both the experimental and model simulation studies are based on Iowa populations. Statistical analysis of MC-simulated data Correlation analysis technique was used to estimate the correlations among the simulated radon, airborne bimodal PAECs, dose rates and surface deposited radon progeny. Multiple linear regression models were used to identify a quantitative relationship between the airborne radon progeny dose rate and the surface-deposited radon progeny. Because the radon concentrations, airborne dose rates, and surface deposited radon progeny activities are all highly right-skewed, natural logarithm

118 100 transformations were applied to all of the above variables so that parametric regression analysis could be used. Cp, adjusted R-square-based model selection, and test-based backward model selection were conducted to choose the best surface-deposited radon progeny or combinations in predicting airborne dose rate. Variables with p-values less than 0.10 were kept in the backward variable selection. Analyses were carried out using SAS statistical software. All analyses were conducted for smoking rooms and nonsmoking rooms separately. Because only binary variable room size was simulated in the MC model as a non-radiation environmental factor, regression analysis was used to evaluate its effect on predicting the dose rate as a covariate. Results The short-term radon distribution in the field study was adopted as the input radon distribution (Figure 5.2) of the MC model, because the airborne PAECs and dose rate were measured simultaneously with short-term radon gas concentration in the field study. The fraction of large rooms 0.62 and fraction of smoking rooms 0.39 from the field study were also adopted as the input parameters of the MC model. In total, 100 data points were simulated by the MC model. Descriptive statistics The descriptive statistics for major physical parameters were summarized in Table 5.3 by smoking status. By chance the MC model produced higher median 222 Rn concentration for nonsmoking rooms as compared to smoking rooms (median 245 versus 146 Bqm -3 ). The medians of MC model output parameters: unattached PAEC, Pdose, Jdose and deposited 218 Po and 214 Po were also higher for nonsmoking rooms as compared to smoking rooms (Table 5.3). However, the median attached PAEC was almost the same between smoking and nonsmoking rooms. The equilibrium factor F and unattached fraction fp, which should have no relation with 222 Rn concentration, were negatively

119 101 related as we expected (Figure 5.3). Smoking was positively related to F and negatively related to fp. Correlation analysis Table 5.4 presents the Spearman correlation coefficients between airborne and deposited variables by smoking status. Nonsmoking rooms In nonsmoking rooms, both deposited 218 Po and 214 Po have high correlations with 222 Rn, airborne unattached PAEC, Pdose, and Jdose (p-value<0.0001). Only the airborne attached PAEC was not so highly correlated with deposited radon progeny. The slow deposition velocity for large size particles could well explain this phenomenon. For airborne dose rate, the correlation between Pdose and the deposited radon progeny is better than that between Jdose and deposited radon progeny, because Pdose puts higher weight on the unattached fraction of PAEC. Smoking rooms In rooms occupied by active smokers, both deposited 218 Po and 214 Po have high correlations with 222 Rn, airborne unattached and attached PAEC, Pdose and Jdose (pvalue<0.0001). Even attached PAEC, which has a smaller deposition rate, presents good correlations with deposited 218 Po and 214 Po in smoking environments due to the large fraction of attached airborne radon progeny. As a result, the correlation between Jdose and deposited radon progeny improved as compared to that in nonsmoking rooms. In both nonsmoking and smoking rooms, the deposited 218 Po and 214 Po were highly correlated to each other (r=0.99, p-value<0.0001). Regression analyses Radon, deposited 218 Po, and deposited 214 Po were considered as potential predictors of airborne dose rate (Pdose and Jdose). In order to achieve a normal

120 102 distribution prior to analyses, the log-scale was used for radon, deposited 218 Po, deposited 214 Po, and airborne dose rate (Pdose and Jdose) in the regression analysis. The selection results based on Cp, adjusted R-square, and backward variable selection methods are presented in Table 5.5. With the exception of what was predicted for Pdose in the nonsmoking environment, the three model selection methods yielded the same results. Although the selection results for Jdose in the nonsmoking environment and Pdose in the smoking environment indicated that the model including radon, deposited 218 Po and deposited 214 Po all together was the best model, further diagnostics of the regression analysis found serious collinearity between deposited 218 Po and deposited 214 Po. The collinearity problem was caused by the high correlation between deposited 218 Po and deposited 214 Po (r=0.99 in Table 5.4), which would seriously impact the accuracy of parameter estimates in the linear regression analysis. So the second best model was selected based on Cp and Adjusted R-square in Table 5.5 with a single deposited radon progeny (deposited 218 Po or deposited 214 Po). This second model was used in the final analysis as shown in Table 5.6 and Table 5.7. Four separate models were built for Pdose and Jdose and by different smoking status. The binary room size variable was tested for significance in each model. The results of the model selection for Pdose are shown in Table 5.6 and the results for Jdose are presented in Table 5.7. For both Pdose and Jdose, under both smoking and nonsmoking conditions, the 222 Rn was a significant positive predictor to dose rate at the 0.05 significance level after adjusting for all other covariates in the model. However, the effect of radon in predicting dose rate was higher for Jdose than for Pdose under both nonsmoking and smoking conditions (RR=1.96 versus 3.05 and RR=2.47 versus 3.06). In addition, under both smoking and nonsmoking conditions, the selected single deposited radon progeny (deposited 218 Po or deposited 214 Po) was noted to be a significant positive predictor for Pdose and a significant negative predictor for Jdose after adjusting for the radon effects.

121 103 For both Pdose and Jdose, the room size variable was only significant in smoking rooms. In smoking environments, after adjusting for the radon and radon progeny effects, a significant increase of 1.14 in the mean Pdose was found for large room sizes relative to small room sizes (95%CI ). Similarly, large room sizes significantly increased the mean Jdose relative to small room sizes. Discussions and Conclusions As compared to the field results in CHAPTER III, the MC simulation study produced lower unattached PAEC (median 0.8 versus 1.6 mwl) and higher attached PAEC (median 22.9 versus 8.1 mwl) as shown in Table 5.3 and Table 3.2. Higher F values (median 0.58 versus 0.22) and lower fp values (median 0.03 versus 0.15) were also observed in MC simulations as compared to the field results. Not surprisingly, the MC simulation study also produced lower Pdose (median 16.0 versus 20.9 msvyr -1 ) and higher Jdose (median 19.1 versus 13.5 msvyr -1 ) as compared to the field results, since Pdose weighs more towards unattached PAEC than attached PAEC. Lower deposited 218 Po and 214 Po concentrations were also noted in MC simulations as compared to the field results (median 7.6 and 6.8 versus 20.4 and 19.2 Bqm -2 ). Overall, the MC model represents an environment that has lower fraction of unattached airborne radon progeny, thus represents a lower depositional environment, as compared to the actual field settings. As we can see from Figure 5.4, in MC simulation study, deposited 214 Po was sensitive to deposition velocity of unattached radon progeny and attachment rate in nonsmoking rooms. Therefore, in MC simulation study, positive shift in the input parameter distribution of deposition velocity or negative shift in that of attachment rate in nonsmoking rooms will increase the output values of deposited 214 Po, which will make the MC simulation values more comparable to the field results. However, some similar characteristics still exist from the field data and MC simulated data after data analysis. For example, in nonsmoking environments, the

122 104 correlation between deposited radon progeny and unattached PAEC was much higher than the correlation between deposited radon progeny and attached PAEC. In fact, the deposition velocity for unattached radon progeny is about 100 times larger than that for attached radon progeny. For another example, for both Pdose and Jdose, room size is a significant positive predictor for dose rate in the smoking environment, but not in nonsmoking environments. The phenomena mentioned in the above examples can be found in both field data and MC simulated data. Because the MC data are free of huge uncertainties that occur in the actual field setting, the MC findings have the potential to discover useful information that otherwise would be submerged in the sea of error in the real world. For example, in the MC study, the deposited radon progeny was a positive predictor in addition to radon in predicting Pdose, but a negative predictor in addition to radon in predicting Jdose. The underlying factor may be that Pdose is weighted more toward unattached PAEC, which more positively relates to deposited radon progeny than attached PAEC. Overall, computer-based MC simulation is a good tool to study radon progeny behavior in the indoor environment. Further refining of the input parameter distributions, according to the information achieved from the actual field study, enables the MC model to better simulate the radon progeny behavior.

123 105 Table 5.1 Model parameter and other input parameter probability distributions Parameters Distribution Distribution parameters Normal Mean S.D. Surface-to-volume ratio (m -1 ) Large room Small room Lognormal Geo. Mean Geo. S.D. Attachment rate (h -1 ) Smoky room Nonsmoky room Deposition velocity for unattached Po (mh -1 ) Ventilation rate (h -1 ) Fraction of time when smoking present for a smoking room Uniform Min Max Large room selector 0 1 Smoking room selector 0 1 Note: Distributions for model parameters represent a summary of the results given in [66, 114, 161, 164]

124 106 Table 5.2 Model constants and dose conversion factors Model Constants Value Recoil from aerosols 0.83 α -recoil of 214 Pb or 210 Pb into glass due to decay of 218 Po or 214 Po 0.5 α -recoil of 210 Pb out of glass due to decay of 214 Po 0.3 Dose Conversion Factors (msvyr -1 WLM -1 ) a DCF for unattached PAEC b DCF for attached PAEC James [79] c Porstendörfer [104] d ICRP [71] Note: a 1 year (yr) = 52 working months (WM) b PAEC=potential alpha energy concentration c The absorption dose was converted to effective dose 1Gy =2.4Sv d Assuming equal amount of nose and mouth breathing

125 107 Table 5.3 Descriptive statistics for MC simulated radon, airborne PAECs, dose rates, F, fp and surface deposited radon progeny by smoking status Variables Environment (n) Range Median Mean (SD) Unit 222 Rn Unattached PAEC Attached PAEC Pdose Jdose F fp Deposited 218 Po Deposited 214 Po Nonsmoking (53) (204.5) Smoking (47) (158.5) Total (100) (186.8) Nonsmoking (53) (2.3) Smoking (47) (0.7) Total (100) (1.9) Nonsmoking (53) (30.1) Smoking (47) (30.2) Total (100) (30.0) Nonsmoking (53) (25.1) Smoking (47) (12.3) Total (100) (21.4) Nonsmoking (53) (20.7) Smoking (47) (17.5) Total (100) (19.5) Nonsmoking (53) (0.18) Smoking (47) (0.11) Total (100) (0.17) Nonsmoking (53) (0.13) Smoking (47) (0.02) Total (100) (0.10) Nonsmoking (53) (21.2) Smoking (47) (5.2) Total (100) (16.9) Nonsmoking (53) (17.2) Smoking (47) (4.5) Total (100) (13.8) Bqm -3 mwl mwl msvyr -1 msvyr -1 Bqm -2 Bqm -2

126 108 Table 5.4 Spearman correlation coefficients and p-values under zero-correlation none hypothesis between airborne and deposited variables by smoking status for MC study Environment Nonsmoking (53) Smoking (47) Airborne variables 222 Rn 0.68 <.0001 Unattached PAEC 0.89 <.0001 Attached PAEC Pdose 0.82 <.0001 Jdose 0.59 <.0001 Deposited radon progeny 218 Po 0.71 < < < <.0001 Deposited 218 Po 0.99 < Rn 0.79 <.0001 Unattached PAEC 0.88 <.0001 Attached PAEC 0.68 <.0001 Pdose 0.80 <.0001 Jdose 0.74 < < < < < <.0001 Deposited 218 Po 0.99 < Po

127 109 Table 5.5 Model selection of log-scale a MC simulated radon and deposited radon progeny as predictors of airborne dose rates (Cp, Adjust-R-squre and backward model selection method were used) Environment Dose being predicted Selection method 222 Rn 222 Rn and deposited 214 Po 222 Rn and deposited 218 Po 222 Rn and deposited 214 Po, 218 Po Nonsmoking Pdose Adj-R Cp Backward b * Jdose Adj-R Cp Backward b * Smoking Pdose Adj-R Cp Backward b * Jdose Adj-R Cp Backward b * Note: a All dependent variables and explanatory variables in the regression analysis are in log-scale. b Variables with p-values less than 0.10 were kept in the backward variable selection and * indicates the best selected model.

128 110 Table 5.6 Estimated relative mean (95% CI) change in Pdose for selected predictors using multiple regression analysis by smoking status for MC study Environment Predictor (n) Relative mean Pdose Nonsmoking Smoking a Reference category Ln 222 Rn (unit 1 increase) (53) 1.96 ( ) Ln Deposited 214 Po (unit 1 increase) (53) 1.46 ( ) Ln 222 Rn (47) 2.47 ( ) (unit 1 increase) Ln Deposited 218 Po (47) 1.10 ( ) (unit 1 increase) Size (47) large (27) 1.14 ( ) small (20) 1.00 a

129 111 Table 5.7 Estimated relative mean (95% CI) change in Jdose for selected predictors using multiple regression analysis by smoking status for MC study Environment Predictor (n) Relative mean Jdose Nonsmoking Smoking a Reference category Ln 222 Rn (unit 1 increase) (53) 3.05 ( ) Ln Deposited 218 Po (unit 1 increase) (53) 0.90 ( ) Ln 222 Rn (47) 3.06 ( ) (unit 1 increase) Ln Deposited 218 Po (47) 0.90 ( ) (unit 1 increase) Size large (27) 1.13 ( ) small (20) 1.00 a

130 Figure 5.1 Schematic representation of the room model 112

131 Figure 5.2 Input radon distribution of Monte Carlo model 113

132 Figure 5.3 The relationship between equilibrium fraction F and unattached fraction fp under MC simulation study 114

133 Figure 5.4 Sensitivity analysis for surface-deposited 214 Po 115

134 116 CHAPTER VI DISCUSSION AND FUTURE WORK In most of traditional radon and lung cancer epidemiologic studies, only contemporary radon gas concentration was measured as a surrogate for the subject s lifetime exposure to alpha radiation dose from radon progeny. For practical reasons, it is nearly impossible to actively measure airborne radon progeny dose in large-scale population-based epidemiologic studies involving radon exposure. However, the usage of glass-based passive detector measuring deposited 218 Po and 214 Po and implanted 210 Po, in addition to radon gas measurement provides a reasonable surrogate measure of dose in an instrument that is rather easy to use. Our current research aimed to investigate how well the contemporary surfacedeposited radon progeny measured by a small inexpensive passive glass-based detector (RRD), combined with contemporary radon gas measurement, could predict the contemporary airborne radon progeny dose rate, and what domestic environmental factors could influence this relationship. The three research aims mentioned in Chapter I: Aim 1) Field calibration of the RRD surface-deposited radon progeny as a predictor of the airborne radon progeny dose rate; Aim 2) Field evaluation of temporal and spatial variations of surface-deposited radon progeny and the airborne radon progeny dose rate; Aim 3) Comparison of the field findings with Monte Carlo model calculations; were achieved in Chapter III, Chapter IV, and Chapter V, respectively. The results of this work will provide a quantitative link from the measured deposited 218 Po and 214 Po and radon gas to the unmeasured airborne radon progeny dose rate under different domestic aerosol and ventilation conditions. Once the relationship between the implanted 210 Po in glass and the deposited 218 Po and 214 Po on glass surface is also thoroughly investigated through both field measurements and laboratory tests, the RRD can be finally calibrated and provide the final resolution of the reanalysis of the

135 117 lung cancer risks for the IRLCS based on radon progeny exposure estimates obtained from RRD measurements. In addition, the results of this study will be used as the basis for a large-scale pooled analysis of the Iowa and Missouri Residential Radon Studies, both of which incorporated the use of the glass-based detectors within their study design. The well calibrated RRD could also be used in future large scale epidemiologic studies involving radon exposure assessment. From the occupational and environmental health perspective, the technique and knowledge developed in this study will benefit the overall methodology of assessing exposure to radon, which is a well recognized lung cancer risk in the indoor domestic environment and an unappreciated risk in the general occupational setting [195]. This dissertation research has several strengths. Firstly, the novel active airborne radon progeny detectors (ARPDs) were used in the actual field measurements to provide an estimate of the airborne radon progeny dose rate by measuring both unattached and attached PAECs. Secondly, the effects of multiple indoor environmental factors were investigated in both the calibration process of RRD and the evaluation of temporal and spatial variations. Thirdly, the field findings were compared to the Monte Carlo model calculations. Therefore, the radon progeny behavior in domestic environments could be investigated both practically and theoretically. The research has also several limitations. Firstly, the current RRD couldn t provide good and consistent deposited 218 Po results, especially in smoking environment. As we saw in Chapter III, 26 out of 98 measurements failed to detect deposited 218 Po, leaving the airborne radon progeny measurements for those 26 rooms largely unused, which seriously reduced the sample size of the whole field study and thus lowered the statistical power of this study to detect more delicate confounding and interaction effects between radon, radon progeny, and environmental factors. Secondly, the glass side of RRD is still undergoing final laboratory calibration. The implanted 210 Po calibration of

136 118 RRD still needs further validation due to the difficulty in accurately accounting for the natural alpha radiation within the glass. Once the implanted 210 Po calibration is finished, the quantitative relationship between surface-deposited deposited radon progeny and implanted 210 Po can be further investigated using the current data set. Glass surface information such as glass type, glass age, cleaning frequency, and dust film are potential environmental factors affecting the above relationship that are also included in the current data set. Examinations of the intrinsic radiation in the glass and explorations of the additional glass surface information, described above, is the next planned research following the dissertation. In addition to the actual field calibration of RRD, further calibration of the RRD in a controlled exposure room, which simulates various residential depositional environments, is a very important follow-up activity. Currently, Dr. Steck has a simulated room-like radon chamber in his lab for which both radon level and environmental factors can be controlled to simulate various residential depositional environments, such as smoking, candle burning, or cooking fuel, humidifier usage, air cleaner usage, ceiling or floor fan usage, etc. Performing additional side-by-side ARPD and RRD comparisons in the radon chamber, under various pre-designed aerosol and ventilation conditions, would also produce invaluable information for comparison with the actual field investigation results. After both field calibration and laboratory validation of the RRD, the next work will be applying the calibration results to the IRLCS data set, since the IRLCS incorporated the use of similar glass-based retrospective radon detectors and also investigated similar questions regarding the residential depositional environments. Thus the resolution of the final reanalysis of the lung cancer risks for the IRLCS based on radon progeny exposure estimates obtained from RRD measurements could be achieved. Furthermore, the RRD-based results of IRLCS could be used in a large-scale pooled analysis of the Iowa and Missouri Residential Radon Studies, since both of them

137 119 incorporated the use of the glass-based detectors within their study design. Thereafter, the lung cancer risks estimated from the measurements of glass-based retrospective radon detectors in the pooled analysis could be compared to the lung cancer risks estimated from the traditional contemporary radon gas measurements to finally answer the question: Does the utility of RRD in addition to year long radon gas measurements provides a better way of assessing radon exposure in large-scale radon and lung cancer epidemiologic studies? If the answer is YES, this technique will provide a better estimate of radoninduced lung cancer risk, which will benefit the scientific knowledge in radiation protection and health physics research fields.

138 120 APPENDIX A FORM SENT TO THE POTENTIAL PARTICIPANTS FOR SCREENING RADON TESTS WITH E-PERMS

139 FORM FOR PRELIMINARY RADON TEST NAME: PHONE: ADDRESS: AGE OF THE HOUSE: YEARS LIVED AT CURRENT HOME: CURRENT SMOKED IN THE HOUSE? YES NO EVER SMOKED IN THE HOUSE? YES NO TEST ENVIRONMENT Was the house closed up during the testing? YES NO Was the AC running during the testing? YES NO Do you have forced air heat? YES NO E-PERM DATA LEVEL E-PERM ID ROOM TYPE INSTALL DATE INSTALL TIME REMOVAL DATE REMOVAL TIME BASEMENT FIRST FLOOR 121

140 122 APPENDIX B QUESTIONNAIRE FOR COLLECTING BOTH RADIATION AND NON-RADIATION DATA IN THE SELECTED HOUSE

141 123 RRD and ARPD Data Sheet House information House ID# Installer's initials: Age of house: Age of home: Number of children living in the home: Date Installed: Date Retrieved:

142 124 RRD and ARPD Data Sheet Room information Room ID# Gas Detector ID# Room type Room dimensions: Ceiling height: House level: ( Basement, First floor, Second floor, Third floor, Other ) Room contents and furnishings => take pictures (Digital Camera) Smoking: Estimate the average number of cigarettes smoked per day in this room in the Present Windows open when smoking? Frequent smoking in nearby room? Past Windows open when smoking? Frequent smoking in nearby room? When did the smoking change or stop? Frequently used aerosol sources: ( None Cooking (Next door) Candles Fireplace (Gas or wood) Air Freshener Room Humidifier Other ) Cooking Fuel: ( Electric Gas ) Frequently used aerosol sinks: ( None Room Air cleaners Open windows Other ) Frequently used AC: ( None Central air Window unit ) Frequently used air movers: ( None Ceiling fans Floor fans Forced air heat/cool Other )

143 125 Current activity log Pre-placement: Smoking o Cigarettes per day in measurement room o Cigarettes per day in adjacent rooms connected via open doors o Cigarettes per day in house Heating o Hours of operation per day o Hours of air circulation operation per day Cooling o Hours of HVAC operation per day o Hours of fan operation per day Room air cleaner operation number of hours per day Window open number of hours per day Post-placement: Smoking o Cigarettes per day in measurement room o Cigarettes per day in adjacent rooms connected via open doors o Cigarettes per day in house Heating o Hours of operation per day o Hours of air circulation operation per day Cooling o Hours of HVAC operation per day o Hours of fan operation per day Room air cleaner operation number of hours per day Window open number of hours per day

144 126 ID Type Age Yrs + Cleaned W,M,Y,I Surface information Film N,L,M,H Wipe 0 to 10 Obstacle Y,N Speeds V/H/N (mps) S1 / / S2 / / Detector ID Type: If ordinary glass, then Cabinet, Door, Mirror, Painting, Window; if not ordinary glass then identify as coated glass, leaded glass, plastic, metal, or wood. Age >5 years, the uncertainty (+) in the surface age should be no greater than 20%. Cleaning frequency: Ask the homeowner about recent window cleaning practices. (Weekly, Monthly, Yearly, Infrequently) Film: Visible coating (None, Light, Medium, Heavy). Where light = barely detectable; heavy = almost opaque (avoid if possible). Wipe: Use an alcohol pad to an area the size of your hand; look at the pad and assign a wipe score from 0=no visible dirt to 10=black. Obstacles: (take picture) Plants, vases, chairs, bottles, curtains, i.e. small objects that might affect air currents or deposition. Speeds: If there are air movers (fans, radiators), make two sets of measurements, one with movers off and the other with movers on. Use puff test to get direction vertically and normal to surface. ARPD data Start Date/Time EPERM ID Start V End V Rn Flow meter Start Lpm End Lpm (goal: 1.5) Hour meter Start End End Date/Time Detectors removed time (goal: 3 hours after end time) Thoron progeny detector exposed from (date/time) to

145 127 Notes: Diagram: Room Diagram showing surface detector locations (S), Rn/Tn gas and EPERM detectors (G and E) windows (W), doors (D), fans (F), HVAC registers (R), and ARPD.

146 128 APPENDIX C THEORETICAL CALCULATIONS OF AIRBORNE PAEC BY BIMODAL SIZE FRACTION BASED ON ARPD TRACK DENSITY RESULTS The working mechanism of ARPD is illustrated in Figure C.1. The pump circulates the air through the ARPD system at a constant flow rate Q m 3 s -1 for time T s, which is usually 3-7 days, large enough for the short-lived radon progeny to reach equilibrium. The air flow first passes through the screen, which collects the unattached fraction of the airborne radon progeny, then passes through the filter, which collects all the remaining particles (aerosol attached fraction of the airborne radon progeny). Both the screen and the filter have an effective area of collecting radon progeny S mm 2. On either CR-39 chip, which is 1 cm above the screen or filter, one region responds to 214 Po and 218 Po alphas (with efficiencies ε14 and ε 18 ) through a thicker plastic x film producing lighter alpha track density D 1 tracks/mm 2, while the other region responds to 214 Po and 218 Po (with efficiencies ε 24 and ε 28 ) through a thinner plastic film x producing heavier alpha track density D 2 tracks/mm 2 x ( D 2 > unattached part and x respectively. D ), where x 1 x = u for the = a for the attached part responding to the screen and filter, The denotations used in the following calculations are either mentioned above or illustrated in Table C.1. In all the following calculations, the denotation x unattached part and x On the screen or filter = a for attached part. = u for dn dt x 1 x N10Q x = λ1n 1 (C.1) S dn dt x 2 x N20Q x x = + λ1n1 λ2n2 (C.2) S

147 129 dn dt N Q x x = + λ2n2 λ3n3 (C.3) S x x 3 30 t, we can get the solutions: N x 1 x QN10 = (C.4) Sλ 1 x x x Q N10 + N 20 N2 = S λ2 x x x x Q N10 + N20 + N 30 N3 = S λ3 On the chip (C.5) (C.6) x x x D1 1ε18 A4ε14 A + = (C.7) T x x x D2 1ε28 A4ε24 A + = (C.8) T x x x x x Solve A 1 and A 4 and link them to A 10, A 20 and A 30 : x x x x x 1 D2ε14 D1ε24 N10Q A10 Q A1 = = = T ε ε ε ε S λs (C.9) x x x x x x 1 D1 ε28 D2ε18 x Q x x x Q A10 A20 A 30 A4 = A3 = ( N10 + N20 + N30 ) = + + T ε 28ε14 ε18ε24 S S λ1 λ2 λ3 (C.10) x x x Then partially solve A 10, A 20 and A 30, we have : A S D ε D ε = λ QT ε ε ε ε x x x x ( ε + ε ) ( ε + ε ) x x x A D D 20 A30 S + = λ λ QT ε ε ε ε (C.11) (C.12) x x x To solve A 20 and A 30 separately, information on the relationship between A 20 and x A 30 is needed. However, without that information, we can still derive the equations for airborne potential alpha energy concentration (PAEC) based on the above equations.

148 130 PAEC and related parameters PAEC PAEC x x x x x 3 A 10 A20 A 30 MeVm = ( ) λ1 λ2 λ3 A A A = λ1 λ2 λ3 x x x / [ mwl] ( ) Combined with C.11 and C.12, we can get: (C.13) PAEC x [ mwl] S = QT x x x x ( D1ε28 D2ε18) + ( D2ε14 D1ε24) ( ε ε ε ε ) / (C.14) x x x x PAEC x2 PAEC x2 PAEC x2 [ ] = x 1 + x 2 + ε x i D1 D2 i ε i x dpaec mwl dd dd d (C.15) i = 18,28,14,24; PAEC in mwl Further experiments in the lab found that the current ARPDs underestimate the PAECs for unknown reason. Side-by-side calibrations using other well calibrated radon progeny detectors yielded a correction factor 1.52 for 2.10 for u PAEC and a correction factor a PAEC. The calibration factors used in the above equations can be found in Table C.2. From the PAEC we can derive the equilibrium factor between radon progeny and radon F and unattached fraction of PAEC f p, as well as the airborne dose rate based on the effective dose conversion factors in Table 2.1: F = 3.74 u a ( PAEC + PAEC )[ mwl] 3 Rn Bqm (C.16) f u PAEC = PAEC + PAEC P u a u a ( ) 1 1 Jdose msvyr WLM PAEC PAEC (C.17) = /1000 (C.18) x PAEC in mwl u a ( ) = /1000 (C.19) x PAEC in mwl 1 1 Pdose msvyr WLM PAEC PAEC

149 131 The uncertainties for the above parameters can also be derived using formula similar to C.15.

150 132 Table C.1 Decay constants and denotations for short-lived radon progeny in ARPD Elements 222 Rn: Po: Pb: Bi: Po: 4 Decay Constants (s -1 ) Particle Density (m -3 ) λ λ λ λ λ 4 = In the Air Activity (Bqm -3 ) On the Screen or Filter Particle Density (m -2 ) Activity (Bqm -2 ) = 6 ~ Rn ~ ~ = 3 x x x x x x N 10 A10 = λ1 N10 N 1 A1 = λ1n 1 = 4 x x x x x x N 20 A20 = λ2n20 N 2 A2 = λ2n2 = 4 x x x x x x N 30 A30 = λ3n30 N 3 A3 = λ3n3 x x x x x x N A40 = λ4n40 N 4 A4 = λ4n4 x x x Note: x N 0, A = A and N x 4 0, A x x 3 = A4 = u for unattached part and x = a for attached part of radon progeny.

151 133 Table C.2 Calibration factors and their uncertainties of ARPD Factor Value Factor Value ε dε ε dε ε dε ε dε S mm 2 x dd D x 1 Correcting for u PAEC Correcting for a PAEC x 1.52 dd D x drn 0.10Rn

152 Figure C.1 The schematic drawing of the working mechanism of ARPD 134