Evaluating for a Geospatial Relationship Between Radon Levels and Thyroid Cancer in Pennsylvania

Transcription

1 The Laryngoscope VC 2014 The American Laryngological, Rhinological and Otological Society, Inc. Evaluating for a Geospatial Relationship Between Radon Levels and Thyroid Cancer in Pennsylvania Neerav Goyal, MD, MPH; Fabian Camacho, MS, MA; Joseph Mangano, MPH, MBA; David Goldenberg, MD, FACS Objectives/Hypothesis: To determine whether there is an association between radon levels and the rise in incidence of thyroid cancer in Pennsylvania. Study Design: Epidemiological study of the state of Pennsylvania. Methods: We used information from the Pennsylvania Cancer Registry and the Pennsylvania Department of Energy. From the registry, information regarding thyroid incidence by county and zip code was recorded. Information regarding radon levels per county was recorded from the state. Poisson regression models were fit predicting county-level thyroid incidence and change as a function of radon/lagged radon levels. To account for measurement error in the radon levels, a Bayesian Model extending the Poisson models was fit. Geospatial clustering analysis was also performed. Results: No association was noted between cumulative radon levels and thyroid incidence. In the Poisson modeling, no significant association was noted between county radon level and thyroid cancer incidence (P 5.23). Looking for a lag between the radon level and its effect, no significant effect was seen with a lag of 0 to 6 years between exposure and effect (P to P 5.59). The Bayesian models also failed to show a statistically significant association. A cluster of high thyroid cancer incidence was found in western Pennsylvania. Conclusions: Through a variety of models, no association was elicited between annual radon levels recorded in Pennsylvania and the rising incidence of thyroid cancer. However, a cluster of thyroid cancer incidence was found in western Pennsylvania. Further studies may be helpful in looking for other exposures or associations. Key Words: Ecological study, radon, thyroid cancer, Pennsylvania, geospatial clustering. Level of Evidence: NA Laryngoscope, 125:E45 E49, 2015 INTRODUCTION One of the few known environmental associations with thyroid cancer is a history of radiation exposure. Ionizing radiation in particular has been associated with a variety of cancers (colorectal, parotid, sarcomas). It is defined as particles (such as x-rays, a-rays, b-rays, c- rays) with enough energy to remove an electron from an atom or molecule, thus creating an ion. These ions and free radicals can cause damage to DNA and in effect initiate tumorigenesis. This effect has an established carcinogenic potential thought to occur in a dose-dependent fashion. Incidents with large releases of ionizing radia- From the Pennsylvania State University Milton S. Hershey Medical Center (N.G., FC., D.G.); Department of Surgery, Division of Otolaryngology Head and Neck Surgery (N.G., D.G.); Department of Public Health Sciences, Division of Health Sciences Research (F.C.); and the Radiation and Public Health Project (J.M.), The Pennsylvania State University, Hershey, Pennsylvania, U.S.A. Editor s Note: This Manuscript was accepted for publication June 4, This work was funded by the 2012 AHNS Alando J. Ballantyne Resident Research Pilot Grant (CORE grant). The authors have no other funding, financial relationships, or conflicts of interest to disclose. Send correspondence to Neerav Goyal, MD, Division of Otolaryngology Head & Neck Surgery, Department of Surgery, The Pennsylvania State University Milton S. Hershey Medical Center, 500 University Drive, Mail Code H091, Hershey, PA ngoyal@psualum.com DOI: /lary tion, such as the atomic bomb sites of Hiroshima and Nagasaki, 1 3 as well as the Chernobyl nuclear plant disaster, 4 8 have shown an increased risk of cancer, including thyroid cancer, in the radiation-exposed populations. Another source of radiation is radon. Radon is a colorless, odorless, tasteless radioactive gas that is a decay product of uranium in the soil. It decays via ionizing radiation with the release of a particles. Additionally, as a gas, it is easily inhaled, and its decay products can layer the aerodigestive tract. More recent ecological studies have evaluated the possible association with environmental exposure to radon and cancer. An ecological study performed in Portugal concluded that 18% to 28% of lung cancer mortality could be attributed to indoor radon exposure after accounting for smoking habits. 9 Additionally, radon was found to be significantly associated with chronic obstructive pulmonary disease mortality in Canada. 10 An ecological study using postal codes in southwest England also found an exposureresponse relationship between residential radon levels and squamous cell carcinoma. 11 Kendall et al. have also described that radon and its byproducts can affect multiple organs in the body via inhalation of the gas as well as by transport of the gas through the blood stream. 12 Although ionizing radiation exposure from a single event (such as Chernobyl) has an established association with thyroid cancer, the majority of patients seen with thyroid neoplasms have no history of ionizing radiation E45

2 exposure. The data released by Pennsylvania and the United States display an unexplained increase in the incidence of thyroid cancer, with a more rapid increase in incidence between 1996 and 2005, than any other cancer site. Additionally, among the states, Pennsylvania has one of the highest incidences and rises in incidences of thyroid cancer. 13 Currently, there is no explanation for this discrepancy in Pennsylvania or the United States. Given the historical nuclear accident in central Pennsylvania with Three Mile Island, recent research has evaluated whether this accident has a correlation with thyroid cancer in the state. These studies did not find an association or causal relationship between the Three Mile Island nuclear reactor and the prevalence of thyroid cancer in the state, though there may be a higher incidence of thyroid cancer than expected in the area Previous literature has shown that ionizing radiation is associated with thyroid cancer. 7,16 Additionally, the ingestion and inhalation of radiation products has been shown to be correlated with thyroid cancer. 16 Given the high rates of radon levels in Pennsylvania and the link between radon exposure (both via inhalation and through the bloodstream) and carcinogenesis, 12,17 we hypothesize that there may be an association between thyroid cancer and radon exposure. This study evaluated the levels of radon evident in Pennsylvania to determine whether any correlations exist between radon and the incidence of thyroid cancer in the state. MATERIALS AND METHODS Institutional review board approval was obtained from The Pennsylvania State University Milton S. Hershey Medical Center for this study. Publicly available deidentified databases as well as thirdparty databases were used for the study. The Pennsylvania Cancer Registry is a publicly available database of new cases of cancer diagnosed or treated within Pennsylvania, beginning in 1985, including data compiled from hospitals, clinics, laboratories, radiation facilities, surgical centers, cancer centers, doctors offices, death certificates, as well as information exchange when Pennsylvania residents are diagnosed or treated in other states. It serves as a representative population for retrospective studies, such as this one, due to its comprehensive nature. Data were collected from the Pennsylvania Cancer Registry, which provided information on each case of thyroid cancer diagnosed in Pennsylvania residents from 1991 to Thyroid cancer was identified by an International Classification of Diseases for Oncology 2 and 3 primary site code of C739. Using crude rates as well as county information for each case obtained through the registry, this information was used to calculate thyroid cancer incidence for each geographic area in Pennsylvania. Radon levels were measured in picocuries per liter of air. Through third-party radon testing agencies, the Pennsylvania Department of Environmental Protection accumulates and aggregates county level data on radon levels. This information is obtained by self-testing kits that are used and sent in by homeowners and businesses in the state. Information was obtained from the agency for the years 1991 to Additional similar information was obtained from a third-party agency (Air Check, Inc., Mills River, NC) for the years 1995 to E46 In performing an analysis of the data, county-level comparisons were made. We began by assuring data reliability by comparing our two sources of radon level data to confirm correlation between the datasets. This was done by performing a Pearson correlation between the two datasets. Additionally, radon measurements from previous years were compared to more recent years via a relative ranking method to determine whether there is consistency between measurements. After assessing the reliability of the data, a variety of models were fit to assess the correlation, if any, between thyroid cancer incidence and radon levels. A Poisson regression model with random effects was fit to the log of the expected thyroid cases for a specific year and county using the dependent variables of radon rate, log total population, year effect, as well as a random variable from a normal distribution to assess county effect. Additional models were created to assess for possible overdispersion in the event of predictors that were not accounted for, as well as to include the previous year s thyroid incidence and a possible lag effect of radon on thyroid cancer. A Bayesian model was also fit to the data to account for possible unreliability in the radon measurements. A final Poisson model was fit to the third-party data to determine if a difference in fit was evident in the two radon datasets. Significance was determined with an acceptable Type I error of a 5.05, with a P <.05. SAS 9.3 (SAS Institute Inc., Cary, NC) was used to perform the statistical modeling. Additional analysis was performed to evaluate for possible clustering of radon concentrations and whether these clusters correlated with clusters of thyroid incidence. The average radon levels from 1995 to 1999 and 2000 to 2004, as well as the average thyroid cancer incidences from 2004 to 2009, were calculated by county. A Moran s I scatterplot was created to assess whether these data are geographically clustered or whether they are randomly dispersed. Then, the two variables were compared to assess whether or not an association exists. Both a 5- and 10-year lag period for an expected effect of radon were tested for. The association between the variables was calculated on a geographic level using bivariate local indicator of spatial association (LISA) tests. 18 RESULTS From our data, 1,263 (94%) observations out of a possible 1,340 observations from 1991 to 2009 were used in the estimation. Those measurements that were missing data regarding location information were excluded. Using a Poisson model, the regression weight for radon was not significant (b , standard error [SE] , P ), suggesting that radon levels are not predictive of thyroid rate increase. Using a negative binomial regression, we tested for overdispersion. Overdispersion in the original model was not considered excessive (Pearson v 2 /degree of freedom 5 1.5, where no overdisperson occurs at 1). Using negative binomial regression, radon was not found to be significant (b , SE , P ). The next model looked at yearly changes in incidence rates, using the previous year s thyroid cases as a predictor. Again, radon regression weight remained nonsignificant (b , SE , P ). Several different models incorporating various lag combinations were fit, where radon values were substituted by corresponding change variables or previous radon values. Using a variety of lag periods between 2

3 TABLE I. Poisson Regression Models of the Effect of Radon Levels on Thyroid Incidence Comparison to Thyroid Incidence b Standard Error P Value Yearly change in radon levels Previous 2-year average of radon levels Previous 3-year average of radon levels Previous 4-year average of radon levels Previous 5-year average of radon levels The Poisson regression was fit to the log of expected thyroid cases with a variable representing the change in radon levels or the average of radon levels. b refers to the coefficient of the radon variable. and 5 years, the Poisson regression model was reevaluated (as shown in Table I). The values approached statistical significance with a lag period of 3 years (P ), but there were no statistically significant results (P to.1089). To adjust for possibly unreliable measurements of radon from year to year, a Bayesian model was created to create a confidence interval for the actual radon value from year to year. As a final analysis, the Poisson regression model was applied to the third-party (Air Check, Inc.) dataset for the years 1995 to Again, data that were missing location data were discarded. A total of 959 of 1,005 observations (95%) were used. Using this dataset, the regression weight for the radon levels was not significant (b , SE , P ). Overdispersion was similar to the original model. In performing our cluster analysis, we found that both radon levels as well as thyroid incidences for the years measured showed statistically significant evidence rejecting the null hypothesis of random dispersion. For radon, the Moran statistic for levels from 2000 to 2004 (Moran s I , P 5.002) suggests moderate clustering, where a Moran s I of 1.00 would be consistent with perfect clustering. Similarly, for thyroid cancer incidence from 2005 to 2009, the Moran statistic (Moran s I , P 5.02) shows moderate clustering. Figures 1 and 2 show a geographic distribution of a univariate LISA analysis for radon and thyroid cancer incidence, demonstrating areas of significant (P <.05) clustering. Figure 1 highlights either isolated counties or a cluster of counties where the radon level is either significantly (P <.05) higher or lower than the surrounding counties. Figure 2 depicts similar information for thyroid incidence. Figure 2 shows evidence of a high thyroid cancer incidence cluster in western Pennsylvania involving Allegheny, Beaver, Lawrence, Armstrong, Butler, and Mercer counties. There is additionally a cluster of lower thyroid cancer incidence involving Snyder, Mifflin, Huntingdon, and Juniata counties (in central Pennsylvania). Following this, bivariate analyses were performed with both a 5- and 10-year lag of the expected effect from radon. For the 5-year latency period, the Moran statistic was not significant (Moran s I , P 5.074); this was also true for the 10-year latency period (Moran s I , P 5.78). DISCUSSION Over the past 30 years, greater attention has been placed on radon levels in the household in light of evidence associating radon with lung cancer. 19 Similarly, research has found links between radon measurements and cutaneous squamous cell carcinoma. 11 Given the associations and causality between radon exposure and cancer, especially lung cancer, we considered an association between radon and thyroid cancer. Using a similar methodology, this study attempted to demonstrate a similar association between radon levels in Pennsylvania and thyroid cancer incidence. Radon, via radioactive decay, releases a particles, which is in contrast to radioactive iodine that releases c- rays. As such, the radioactive particles from radon decay are heavier and penetrate soft tissues to a lesser degree than c-rays. However, radon exists in a gaseous form, and its decay particles deposit in the airway, and as such could subject the thyroid to additional radiation Fig. 1. Geographic clustering of high radon levels in the northern part of Pennsylvania (McKean, Potter, Cameron, Clinton, and Lycoming counties). There is a cluster of low radon levels in the southwest (Washington, Westmoreland, Somerset counties). There are some isolated counties with significantly lower or higher radon levels than their neighboring counties. E47

4 Fig. 2. Geographic clustering of higher thyroid incidence in the western part of Pennsylvania (Mercer, Lawrence, Butler, Armstrong, Beaver, and Allegheny counties). There is a cluster of low thyroid incidence in central Pennsylvania (Huntingdon, Mifflin, Snyder, and Juniata counties). There are some isolated counties with significantly lower or higher thyroid incidence than their neighboring counties. exposure. Additionally, there is evidence that radon and its byproducts are also carried in the blood stream and found in tissues other than the respiratory tract, including the kidneys, brain, and liver. 12 The research did not demonstrate a correlation between radon levels and the rise in thyroid incidence in Pennsylvania. Even after accounting for possible variability in radon measurements from year to year, and also considering the possibility of a lag period between the radon levels and a measurable difference in thyroid cancer, no statistically significant correlation could be measured. Current methodologies to measure radon levels involve end users placing self-test kits in their homes. With this method, there is the possibility of erroneous measurements. Using a Bayesian model, we attempted to correct for this by evaluating and correcting for any significant variability in the radon measurements available to us. Notwithstanding, we found that the levels were not significantly correlated with thyroid cancer incidence. The cluster analysis demonstrated that both thyroid cancer incidences as well as radon levels show evidence of clustering within Pennsylvania. Although the two variables did not have a statistically significant correlation with regard to their clusters, it is interesting to note the area of high thyroid cancer incidence in the western portion of the state near Pittsburgh. A similar cluster was not evident around Philadelphia (southeastern Pennsylvania), the other major metropolitan area within the state. This would suggest that this cluster would not be fully accounted for by referral patterns into the city and the associated tertiary care centers. One of the limitations of this study is that this study was performed retrospectively and could only be performed on the data available from 1991 onward. It is possible that the impact of radon on thyroid incidence has a lag period longer than what could be measured with our dataset. Our measurements are based on thirdparty data that are collected by individuals usually the owners of the residence. These measurements are usually not taken at a regular interval, and instead are usually taken at the time of selling a house (though not required) or whenever the homeowner chooses to check a measurement. This can be assumed to be a random sampling of the county radon levels. However, although the method of measurement is somewhat standardized, it too can be subject to some variables in how the radon is collected by the measurement apparatus. Additionally, it is difficult to predict the overall exposure of an individual to radon, as measurements are classically made at an individual s residence, and thus does not take into account exposures from the workplace. CONCLUSION Our data do not demonstrate a statistically significant correlation between historical radon levels and the historical rise in thyroid incidence. Although there are some limitations to the dataset utilized, statistical methodology to correct for variability in measurements did not help to show a relationship. Despite there being high levels of radon present in various Pennsylvania counties, this did not match with similarly high thyroid incidence. Clustering analysis shows some interesting findings regarding a cluster of higher thyroid cancer incidence in western Pennsylvania. Further studies evaluating this cluster may yield possible environmental or confounding factors to explain this rise. Additionally, further research evaluating radon through sources, such as the water supply, may yield more information regarding a possible association with thyroid cancer. Acknowledgments The authors acknowledge the American Head and Neck Society and the Centralized Otolaryngology Research Efforts Grants Program for grant support through the 2012 AHNS Alando J. Ballantyne Resident Research Pilot Grant. BIBLIOGRAPHY 1. Nakachi K, Hayashi T, Hamatani K, Eguchi H, Kusunoki Y. Sixty years of follow-up of Hiroshima and Nagasaki survivors: current progress in molecular epidemiology studies. Mutat Res 2008;659: Richardson DB. Exposure to ionizing radiation in adulthood and thyroid cancer incidence. Epidemiology 2009;20: Furukawa K, Preston D, Funamoto S, et al. Long-term trend of thyroid cancer risk among Japanese atomic-bomb survivors: 60 years after exposure. Int J Cancer 2013;132: Ivanov VK, Gorski AI, Maksioutov MA, et al. Thyroid cancer incidence among adolescents and adults in the Bryansk region of Russia following the Chernobyl accident. Health Phys 2003;84:P46 P Nikiforov YE. Radiation-induced thyroid cancer: what we have learned from Chernobyl. Endocr Pathol 2006;17: Jacob P, Bogdanova TI, Buglova E, et al. Thyroid cancer risk in areas of Ukraine and Belarus affected by the Chernobyl accident. Radia Res 2006;165:1 8. E48

5 7. Pacini F, Vorontsova T, Demidchik EP, et al. Post-Chernobyl thyroid carcinoma in Belarus children and adolescents: comparison with naturally occurring thyroid carcinoma in Italy and France. J Clin Endocrinol Metab 1997;82: Kesminiene A. Evrard AS, Ivanov VK, et al. Risk of thyroid cancer among Chernobyl liquidators. Radiat Res 2012;178: Veloso B, Nogueir, JR, Cardoso MF. Lung cancer and indoor radon exposure in the north of Portugal an ecological study. Cancer Epidemiol 2012;36:e26 e Turner MC, Krewski D, Chen Y, Pope CA III, Gapstur SM, Thun MJ. Radon and COPD mortality in the American Cancer Society Cohort. Eur Respir J 2012;39: Wheeler BW, Allen J, Depledge MH, Curnow A. Radon and skin cancer in southwest England: an ecologic study. Epidemiology 2012;23: Kendall GM, Smith TJ. Doses to organs and tissues from radon and its decay products. J Radiol Prot 2002;22: Goyal N, Camacho F, Mangano J, Goldenberg D. Thyroid cancer characteristics in the population surrounding Three Mile Island. Laryngoscope 2012;122: Levin RJ. Incidence of thyroid cancer in residents surrounding the Three Mile Island nuclear facility. Laryngoscope 2008;118: Levin RJ, De Simone NF, Slotkin JF, Henson BL. Incidence of thyroid cancer surrounding three mile island nuclear facility: the 30-year follow-up. Laryngoscope 2013;123: Mays CW. Cancer induction in man from internal radioactivity. Health Phys 1973;25: Gerusky TM. The Pennsylvania radon story. J Environ Health 1987;49: Anselin L. Local indicators of spatial association LISA. Geograph Anal 2010;27: National Research Council. Committee on Health Risks of Exposure to Radon. Health Effects of Exposure to Radon. Washington, DC: National Academies Press; E49