Radon is a radioactive gas produced by decay of radium-226, which is widely distributed in uranium-rich soils and granite bedrock.1 A key factor affecting population radon exposure is therefore the underlying geology of the environment. The World Health Organization2 and the International Agency for Research on Cancer3 list radon as a human carcinogen, with an increased risk of lung cancer. Radon may be a causal factor in 3.3% of all lung cancer deaths in the United Kingdom, although most of these are due to a combination of radon exposure and smoking.4 Radon has also been implicated as a risk factor for other cancers, but studies are limited, and the UK Health Protection Agency concluded in a 2009 review that “Overall there is no epidemiological evidence to suggest that radon exposure contributes directly to excess disease or mortality other than of lung cancer.”5 (p. 42)
Despite the lack of strong epidemiologic evidence, theoretical dosimetry models indicate that radon at average UK exposure levels could be responsible for a proportion of skin cancers. Also, an earlier ecologic study in the United Kingdom suggested an association with nonmelanoma skin cancer.6–8 Because radon and its decay products are attracted to water molecules and some atmospheric gases, it is possible that the resulting aerosols could adhere to the skin via electrostatic attraction, leading to prolonged irradiation of the skin, even when the person is no longer in areas of high exposure.9
Skin cancers (malignant melanomas and nonmelanomas) are by far the most common form of cancer in the United Kingdom and other countries. Although less prevalent than nonmelanomas, malignant melanoma accounts for the majority of skin cancer mortality in the United States and Europe, with the lifetime risk of an individual developing this disease of approximately 2%.10,11 A key risk factor for all skin cancer types is exposure to ultraviolet radiation, although different modes and intensities of exposure seem to be important for different tumor types.12–14 Ionizing radiation has also been implicated as a risk factor in the development of basal cell carcinoma.12,15,16 Southwest England has the highest melanoma incidence and mortality rates, as well as the highest incidence of nonmelanoma skin cancer, in the United Kingdom.17,18 The southwest region of England also encompasses localities where background levels of radon can far exceed the national average of 21 Bq/m3 (Becquerels per cubic meter).4,19 A previous ecologic study in this region found an association between radon concentration and nonmelanoma skin cancer registrations from 1989 to 1992.6
Completeness of nonmelanoma skin cancer registration is variable nationally18 and internationally.20 However, the South West Public Health Observatory (which hosts the regional cancer registry) leads on skin cancer for the UK Association of Cancer Registries,21 and is recognized for high-quality nonmelanoma surveillance.22 Three key factors make this an ideal region for an updated, improved ecologic study to investigate the association between radon exposure and the risk of developing different types of skin cancer: (1) a high-quality surveillance system providing more complete detection of nonmelanoma skin cancers than is available in many other locations nationally and internationally, (2) the availability of data on skin cancer subtypes, and (3) the wide variation in radon concentrations, permitting comparison of rates from very high to very low radon areas.
METHODS
Geography
The study area included the counties of Devon, Cornwall, and Isles of Scilly in southwest England. Radon and skin cancer registration data were available by postcode sector and ward. However, ward boundaries change over time, and the 2 datasets were not consistently coded. Therefore, we selected postcode sectors, which are aggregations of unit (full) postcodes. Sectors could be aggregated into larger spatial units, but represent a useful compromise between smaller units, which can result in very small populations and hence unstable rates, and larger units, which may lack sufficient resolution to detect geographical variation.
Skin Cancer Incidence
Registration data were obtained with appropriate ethical and data protection permissions. Although this study only involved secondary analysis of routinely collected data, ethical approval was sought and obtained from the South West I Research Ethics Committee (reference 08/H0203/6). Data were stored and analyzed in compliance with local governance requirements.
These registration data included total counts of new skin cancer cases diagnosed during the 5 years between 2000 and 2004 and recorded at the regional cancer registry. Additional information regarding persons diagnosed included age at diagnosis, postcode sector, and tumor type. The 4 tumor types were malignant melanoma plus 3 types of nonmelanoma skin cancers: squamous cell carcinoma, basal cell carcinoma, and “other.” Given the potential differential relationship between radon and various skin cancers,6 we calculated incidence rates separately for the tumor types.
Data included only the first instance of a tumor/lesion recorded for each person for nonmelanoma skin cancer, but multiple instances per person for malignant melanoma. Due to the mismatch between postal, administrative, and census boundaries, and the use of some superseded postcodes, registrations for 7 postcode sectors could not be matched with population data and were excluded. For comparative purposes, we obtained national (England) registration rates for melanoma and nonmelanoma skin cancer.18 National figures also included multiple instances of malignant melanoma per person, and were therefore comparable with study area data, although national nonmelanoma statistics are not subdivided by tumor type.
To calculate incidence rates, postcode sector population estimates for the years 2000–2004 were required. Population counts for postcode sectors were available from 2001 Census Key Statistics tables,23 but estimates for other years were not available. Annual age-specific population growth rates relative to 2001 were therefore calculated from regional population estimates.18 We applied these rates to 2001 postcode sector populations to estimate total person-years at risk from 2000 through 2004. Key Statistics tables did not include population by both age and sex, and therefore incidence rates could not be standardized for both. Consequently, we calculated directly age-standardized rates for postcode sectors, using the total study population as the reference population.
Radon Exposure
We obtained mean indoor radon concentrations (in Bq/m3) for each postcode sector from a National Radiologic Protection Board (now part of the Health Protection Agency) radon atlas.24 These data were accumulated between approximately 1980 and 2000 through government-funded surveys and measurements requested by householders and landlords. The data were collected from approximately 400,000 households of around 22 million in England and Wales. However, measurements are not geographically representative, and surveys were targeted at areas where high levels were expected. Householders/landlords in areas with high radon levels would also be more likely to request measurement (often during housing transactions).24
Radon measurements were made using standardized methods. Two passive radon detectors were placed, one in the main living area and one in a frequently used bedroom. After 3 months, detectors were returned for analysis, producing an average annual radon concentration for the dwelling.25 The atlas takes these measurements and summarizes them for postcode sectors, and the summary mean radon concentration for each sector forms our small-area exposure estimate. Radon measurements within a given area tend to be approximately log-normally distributed, and it has been estimated that the coefficient of variation (standard deviation/mean) within any given small area is 19%.5 This suggests that the mean observed within each sector can be considered to be reasonably representative of households within that area.
Mean radon concentrations across postcode sectors derived from atlas data were classified and analyzed categorically to allow for nonlinear association with skin cancer incidence. We adopted a classification that maintains consistency with previous work,6 dividing the range of values into 10 classes.
Statistical Methods
Age-specific malignant melanoma and nonmelanoma skin cancer registration rates for the study area were compared with national rates. We mapped radon concentrations and directly age-standardized skin cancer incidence rates for visual inspection of geographic variation. Poisson regression models were used to calculate age/sex-adjusted rate ratios across categories of radon concentrations before and after adjustment for the effects of other potential confounders, as follows.
Ultraviolet (UV) radiation from sunlight is a potential confounder if it is independently related to radon exposure. No data were available on actual population exposure, but as a proxy, small-area estimates of daily hours of bright sunshine were obtained from the UK Met Office.26,27 These estimates were available for 5-km grid squares for 1961–1990. We calculated mean daily hours of bright sunshine from April to September during this period for postcode sectors, using a Geographic Information System (GIS).
The following potential sociodemographic confounding variables were derived from 2001 Census tables.23 Given the lack of population denominators by both age and sex, models were constructed using age-specific registration count as the outcome variable, with population as the offset. To allow for variation in sex distribution, we included as an independent variable the percentage of the postcode sector population that was male. Three socioeconomic variables were also included: low socioeconomic status (percentage of people aged 16–74 years who are in National Statistics Socio-economic Classifications 5–7), unemployment (percentage of the economically active population that was unemployed), and low educational attainment (percentage of people aged 16–74 without qualifications). To assess effects of occupational sun exposure, the percentage of the employed population aged 16–74 working in primarily outdoor industries (agriculture, hunting, forestry, fishing, and construction) was also included. Coastal residence may be associated with an increased risk of skin cancer due to increased time spent on beaches and consequent increase in sun exposure.28 Given the extensive coastal zone in the study area, this was an important consideration. The GIS was used to calculate whether each postcode sector was wholly or partially within 2 km of the coast, and this binary indicator was included in regression models. Population density (from 2001 Census) was included to allow for potential confounding effects of urban-rural status.
Poisson regression models were constructed using Stata version 10 (Stata Corp, TX). We calculated robust standard errors (SEs), and hence confidence intervals (CIs), for rate ratios (RRs) to allow for clustering of age-group-level observations within postcode sectors.
RESULTS
Exclusion of 7 study area postcode sectors meant that 44 of the 18,350 skin cancer registrations were not considered. The remaining 287 postcode sectors had a mean population of 4610 in 2001. Age-specific registration rates of malignant melanoma and nonmelanoma skin cancer for the study area and the whole of England are presented in Table 1. These data demonstrate high registration rates of malignant melanoma and nonmelanoma skin cancer in the study area relative to England, for all ages. In the study area, 70% of nonmelanomas were basal cell carcinomas, 24% squamous cell carcinomas, and 6% other nonmelanomas. eTables 1 and 2 (https://links.lww.com/EDE/A536) describe age-specific rates and registration counts for nonmelanoma subtypes. Because the category of other nonmelanoma had small numbers and included tumors of mixed pathology, this group was not considered further.
;)
Table 1: National and Study Area Skin Cancer Age-specific Registration Rates per 100,000 Population per Year, 2000–2004
Given the high radon levels in parts of the study area, radon had been measured in a relatively high proportion of dwellings in this region. Of an estimated 688,389 dwellings in the 287 postcode sectors, 162,634 (24%) had contributed a measurement (compared with 2% nationally). The proportion of dwellings contributing a measurement within study area sectors varied between 3% and 61%. The minimum number of dwelling measurements for a single postcode sector was 14 and the maximum 1700, with only 9 sectors having fewer than 50 measurements. Mean radon concentration across all sectors was 98.1 Bq/m3 (standard deviation=73.1 Bq/m3). Within the study area, sectors with high radon levels tended to have a higher proportion of dwellings measured (correlation coefficient=0.63).
Radon concentrations across the study area were mapped using the Etherington et al classification6 (Fig. 1). Substantial variation in mean radon concentrations can be observed, with lower values in eastern, northern, and southwest Devon, and high concentrations through central Devon and Cornwall and in western Cornwall. The highest mean postcode sector household radon concentration was 475 Bq/m3. The UK Government “action level,” above which remedial action is recommended, is 200 Bq/m3.29
Figure 1: Map of radon concentrations across the study area: mean of value for all dwellings that have had a radon measurement within each postcode sector. Note: Boundary data are © Crown Copyright and supplied by EDINA UKBORDERS with the support of the ESRC and JISC.
Age-standardized rates of malignant melanoma, basal cell carcinoma, and squamous cell carcinoma are shown in Figure 2. Melanoma rates were higher in coastal areas, basal cell carcinoma rates were higher in eastern Cornwall and southern Devon, and squamous cell carcinoma rates were higher in central/western Cornwall. The distributions were all quite different, with only weak correlation coefficients between standardized rates of the 3 types of skin cancer (melanoma vs. basal cell carcinoma=−0.10; melanoma vs. squamous cell carcinoma=−0.10; basal vs. squamous cell carcinoma=−0.03). The variation in geographic patterns supported our decision to analyze the 3 tumor types separately.
Figure 2: Directly age-standardized registration rate per 100,000 population per year, 2000–2004, by postcode sector, of (A) malignant melanoma, (B) basal cell carcinoma, and (C) squamous cell carcinoma. Note: Boundary data are © Crown Copyright and supplied by EDINA UKBORDERS with the support of the ESRC and JISC.
Poisson regression results for the association between mean radon concentration and malignant melanoma registration rate, before and after mutual adjustment for potential confounders are presented in Table 2. No evidence of an association between radon concentration and melanoma incidence was found, nor was evidence of an association with mean daily sunshine exposure found. There was some indication of a higher melanoma risk with proximity to the coast, although statistical evidence was weakened after adjustment for other variables (RR=1.14 [95% CI=0.99–1.31]). There was also evidence of inverse associations of melanoma incidence with the proportions of the population in low socioeconomic classes and with those working in outdoor industries.
Table 2: Association of Population Characteristics With Postcode Sector Malignant Melanoma, Basal Cell Carcinoma, and Squamous Cell Carcinoma Registration Rate, 2000–2004, Based on Poisson Regression Analysis (n=287 Postcode Sectors)
Unadjusted and adjusted rate ratios for basal cell carcinoma by radon levels also fluctuated around the null (Table 2), providing no evidence of an association. Again, there was no suggestion of an association between mean daily sunshine and incidence of basal cell carcinoma. Results from the fully adjusted model indicated inverse associations of basal cell carcinoma incidence with both population density and the proportion of population in primarily outdoor occupations.
In contrast to the other cancer types, squamous cell carcinoma rates clearly increased with higher radon concentrations. This association was almost unchanged after adjustment for potential confounders (Table 2). Incidence rates in areas with the highest radon concentrations (>230 Bq/m3) were 1.76 (95% CI=1.46–2.11) times those in areas with the lowest radon concentrations (0–39 Bq/m3). An association was also observed for squamous cell carcinoma with coastal proximity, but not with mean daily sunshine or any of the socioeconomic factors.
Rate ratios and associated 95% confidence intervals across radon categories for each of the 3 tumor types are shown graphically in Figure 3. This is suggestive of an exposure-response gradient for the association with squamous cell carcinoma. Expanded regression results, including stratum-specific absolute rates and P values for trend, are provided in eTables 3–5 (https://links.lww.com/EDE/A536).
Figure 3: Rate ratios across radon categories from fully adjusted models (
Tables 1,
2). Reference category, 0–39 Bq/m
3 (RR=1). A, malignant melanoma, (B) basal cell carcinoma, (C) squamous cell carcinoma.
DISCUSSION
Comparison of data from the study area with national data supports the assertion that rates of melanoma and nonmelanoma skin cancer are high in southwest England, with crude incidence rates in the study area of almost double the national rates. Rates exhibited strong age gradients, and the older population demographic of the study area may to some extent explain this difference in crude rates.
Using Poisson regression models, incidence rates of the 3 skin cancer types were associated with different environmental and sociodemographic area-level risk factors. If these associations reflect actual individual-level risk factors, they go some way to explaining the varying geographic patterns observed in the maps in Figure 2. Adjusted rate ratios across radon categories provided evidence of an exposure-response relationship between radon and squamous cell carcinoma incidence. Although this finding is subject to study limitations, it provides an important indication that radon may indeed be a risk factor for this type of skin cancer, and warrants further investigation. This finding is consistent with a previous ecologic study carried out in the same area using cancer registration data for 1989–1992.6 This previous study investigated associations between mean small-area radon concentrations and registration rates of a range of cancers, and found an association only with nonmelanoma skin cancer. The authors neither analyzed nonmelanoma subtypes nor published rate ratios. However, comparing published age-standardized rates for 1989–1992 using the same radon categories as in our analysis (>230 vs. 0–39 Bq/m3) produces rate ratios of similar magnitude (1.73 for men, 1.66 for women).
The positive association between radon and squamous cell carcinoma risk is biologically plausible. Both basal and squamous cell carcinomas arise from metastatic transformation of the same cell type (keratinocyte). Squamous cell carcinoma develops within the epidermal skin layer, the thickness of which varies between 70 and 120 μm,30 whereas the progenitor cells of basal cell carcinoma arise from the much deeper intrafollicular epidermis.31,32 Alpha particles emitted by radon or its daughter products can travel 40–70 μm in tissues,33 suggesting that it is able to affect cells only within the epidermal skin layer. Therefore, radon is unlikely to play a role in the transformation of basal cell carcinoma progenitor cells. Furthermore, squamous cell carcinoma occurs predominantly on UV- (and radon-) exposed skin of the head, neck, and back of the hands, whereas a high proportion of basal cell carcinoma occurs on non–sun-exposed areas of skin where clothing also prevents the penetration of alpha radiation.15
The observed association between radon and squamous cell carcinoma incidence could also be due to residual or unmeasured confounding. For example, the analysis adjusts for geographic variation in bright sunshine, and coastal proximity accounts to some extent for the possibility of increased sun exposure on beaches. However, data were not available to adjust for human behavior, which may vary geographically. Because the relationship between frequency/intensity of sunlight exposure and risk of skin cancer varies according to skin cancer type, and behavior resulting in sun exposure may vary geographically, it is plausible that the differential geographical patterns observed could be driven by UV exposure variation. The limited variation in solar irradiance across the relatively small study area (mean bright summer sunshine=6.1 hours/day; range=5.5–6.9 hours/day) means that UV exposure variation will be largely dictated by behavior, which we cannot control for. It is therefore perhaps unsurprising to find no association between the summer sunshine variable and skin cancer registration rates.
Regarding the other associations detected, it is interesting to note a coastal effect for both melanoma and squamous cell, but not basal cell carcinoma; this coastal variable might in fact provide some degree of adjustment for behavior-related UV exposure. Coastal residents may spend more time outdoors with unprotected skin, thereby increasing UV exposure, and a lack of association with basal cell carcinoma might be explained by the frequency/intensity of exposure variation described earlier. The lower incidence of basal cell carcinoma observed in areas with more people in primarily outdoor occupations is not consistent with previous findings, although the lower incidence of melanoma in these areas is as expected.34,35 This inconsistency may indicate that the measure used of outdoor occupational exposure is too crude, and could be subject to residual confounding.
The lower malignant melanoma incidence in areas with more people in lower socioeconomic strata is consistent with previous studies.36,37 Acute, intense exposure to sunlight is associated with increased malignant melanoma risk.14 The higher incidence in higher socioeconomic status populations (and those with greater prevalence of indoor occupations) may therefore indicate generally low UV exposure punctuated by high-exposure episodes, perhaps during holidays in high-UV environments.
Strengths and Limitations
Our investigation uses population data and is based on sufficiently large numbers to detect variations in incidence across the study area. The regional cancer registry has a well-established system for registration of nonmelanoma skin cancers, which is not the case in many areas. The study area includes a very wide range of average radon concentrations, giving the opportunity to compare various levels of exposure.
A key limitation inherent in this study design is the potential for the ecologic fallacy to be in play. Associations observed at the aggregate level (postcode sector) may not necessarily hold at the level of the individual. Thus, although we observe an association between area mean radon concentration and area squamous cell carcinoma incidence, it is possible that people living in dwellings with high-radon concentrations are not the ones who go on to develop this form of skin cancer. Further, as indicated elsewhere, it is possible that the observed association is due to unmeasured or residual confounding. Certain socioeconomic circumstances or behaviors (such as dwelling characteristics or ventilation habits) could affect household radon and also be associated with the outcomes of interest, although we do not have any evidence that this is the case. There is no specific reason for indoor radon to be related to sun exposure behavior, although if that is the case this would be a key unmeasured confounder, given the strength of association between UV and skin cancers. Although these issues cannot be rectified within the confines of this study, the specificity of the association between radon and squamous (and not basal) cell carcinoma lends credence to the findings, given the biologic plausibility and theoretical dosimetry already described.
We assume that our estimate of radon concentration for each postcode sector is a good estimate of radon exposure for all residents within that area, and that household radon concentration is a good indicator of personal exposure. As indicated earlier, it is likely that dwellings within a sector will have variable radon concentrations, depending on immediate geology and building characteristics, although the mean is likely to be reasonably representative of dwellings within the sector.5 Areas with high radon levels tend to have had more measurements made, and so the precision of sector concentration estimates increases with increasing radon concentrations. However, most of postcode sector means are based on a large number of measurements. A sensitivity analysis of the squamous cell carcinoma model, excluding the 9 postcode sectors with fewer than 50 radon measurements, resulted in negligible differences compared with the reported findings. Residents spend differing amounts of time indoors exposed to particular radon concentrations, and so individual radon exposures within dwellings may also vary. However, studies comparing household and personal radon monitoring indicate that dwelling measurements provide a good indication of personal exposure. One study found a moderate correlation (r2=0.52) between personal and dwelling radon measures,38 while another found a strong correlation (r2=0.85).39
A final point regarding exposure estimation is that exposure and outcome measures are for the same point in time, and we therefore assume current exposure (based on long-term accumulated radon measurements for current residence location) is a reasonable estimate of total exposure. This may lead to misclassification, for example where a person classified as living in a high-radon area moved there recently from a low-radon area.
In terms of data issues, postal boundaries are subject to change over time, leading to errors in matching radon and skin cancer data. The “distance-to-coast” indicator provided a reasonable indicator of coastal proximity, but as some postcode sectors are large and extend some way inland, a portion of the population classified as living within 2 km of the coast will have been misclassified. As with many small-area studies, population denominator data that exactly match registration (numerator) data are unavailable. The application of regional population growth rates to postcode sector populations for 2001 presumes that growth/shrinkage was uniform across the whole study area between 2000 and 2004. This is unlikely, and it is possible that if a sector experienced substantial changes in population during this time, the incidence rate for that sector could be an under- or overestimate. However, overall geographical patterns of incidence rates across the study area are unlikely to be strongly affected, and adjustment improves the validity of the comparison of study area rates with national figures.
Many of these limitations are common to ecologic environmental epidemiologic studies, and while they are important, they do not negate our findings. We have found a population-level association between radon exposure and skin cancer incidence, specifically with squamous cell carcinoma. The use of population data means that results should be generalizable to elsewhere in the United Kingdom and other countries. The radon concentrations at which increased risk is observed are also found elsewhere, indicating that this issue is not confined to the study region alone. Further investigation is warranted, particularly to assess individual long-term environmental radon exposure and subsequent risk of squamous cell carcinoma.
ACKNOWLEDGMENTS
Radon data were provided by the Health Protection Agency. Skin cancer data were supplied by the South West Public Health Observatory. Population data source: 2001 Census (Crown copyright 2004). Digital map boundaries use data provided through EDINA UKBORDERS with support of ESRC and JISC and use boundary material, which is copyright of the Crown. Sunshine data adapted from Crown copyright data supplied by the Met Office. Analysis and interpretation of the data are solely the responsibility of the authors and not the data suppliers. We thank 2 anonymous referees for helpful comments and suggestions for revision of the paper.
REFERENCES
1. Health Protection Agency. HPA-RPD-001—Ionising Radiation Exposure of the UK Population: 2005 Review. Available at:
http://www.hpa.org.uk/Publications/Radiation/HPARPDSeriesReports/HpaRpd001. Accessed 24 August 2011.
2. World Health Organization. Tobacco or Health: A Global Status Report. Geneva: WHO; 1997.
3. International Agency for Research on Cancer. Ionizing radiation. Part 2: some internally deposited radionuclides. IARC Monogr Eval Carcinog Risks Hum. 2001;78.
4. Gray A, Read S, McGale P, Darby S. Lung cancer deaths from indoor radon and the cost effectiveness and potential of policies to reduce them. BMJ. 2009;338:a3110.
5. Advisory Group on Ionising Radiation. Radon and public health. Chilton, United Kingdom: Health Protection Agency; 2009.
6. Etherington DJ, Pheby DF, Bray FI. An ecological study of cancer incidence and radon levels in South West England. Eur J Cancer. 1996;32A:1189–1197.
7. Eatough JP, Henshaw DL. Radon dose to the skin and the possible induction of skin cancers. Radiat Prot Dosim. 1991;39:33–37.
8. Henshaw DL, Eatough JP. The theoretical risk of non-melanoma skin cancer from environmental radon exposure. J Radiol Prot. 1995;15:45–51.
9. Eatough JP. Alpha-particle dosimetry for the basal layer of the skin and the radon progeny 218-Po and 214-Po. Phys Med Biol. 1997;42:1899–1911.
10. Hocker TL, Singh MK, Tsao H. Melanoma genetics and therapeutic approaches in the 21st century: moving from the benchside to the bedside. J Invest Dermatol. 2008;128:2575–2595.
11. Meyle KD, Guldberg P. Genetic risk factors for melanoma. Hum Genet. 2009;126:499–510.
12. Ramirez CC, Federman DG, Kirsner RS. Skin cancer as an occupational disease: the effect of ultraviolet and other forms of radiation. Int J Dermatol. 2005;44:95–100.
13. Leiter U, Garbe C. Epidemiology of melanoma and nonmelanoma skin cancer—the role of sunlight. Adv Exp Med Biol. 2008;624:89–103.
14. de Gruijl FR. Skin cancer and solar UV radiation. Eur J Cancer. 1999;35:2003–2009.
15. Dessinioti C, Antoniou C, Katsambas A, Stratigos AJ. Basal cell carcinoma: what's new under the sun. Photochem Photobiol. 2010;86:481–491.
16. Lichter MD, Karagas MR, Mott LA, Spencer SK, Stukel TA, Greenberg ER. Therapeutic ionizing radiation and the incidence of basal cell carcinoma and squamous cell carcinoma. The New Hampshire Skin Cancer Study Group. Arch Dermatol. 2000;136:1007–1011.
17. Office for National Statistics. Cancer Atlas of the United Kingdom and Ireland 1991–2000. Available at:
http://www.ons.gov.uk/ons/rel/cancer-unit/cancer-atlas-of-the-united-kingdom-and-ireland/smps-no–58/chapter-14–melanoma-of-skin.pdf. Accessed 31 October 2011.
18. Office for National Statistics. Cancer Statistics: Registrations Series MB1. Available at:
http://www.ons.gov.uk/ons/taxonomy/index.html?nscl=Health+and+Social+Care. Accessed 31 October 2011.
19. National Radiological Protection Board. Living with Radiation. Oxford: Health Protection Agency; 1998.
20. Curado MP, Edwards B, Shin HR. Cancer Incidence in Five Continents. Vol 9. IARC Scientific Publications No 160. Lyon: IARC; 2007.
21. South West Public Health Observatory. Skin Cancer Hub. Available at:
http://www.swpho.nhs.uk/skincancerhub. Accessed 24 August 2011.
22. South West Public Health Observatory. Non-Melanoma Skin Cancer: Estimates of cases. Available at:
http://www.swpho.vnhs.uk/skincancerhub/resource/item.aspx?RID=52794. Accessed 24 August 2011.
23. Office for National Statistics. Census 2001 Key Statistics for postcode sectors in England and Wales. Available at:
http://www.ons.gov.uk/ons/guide-method/census/census-2001/index.html. Accessed 31 October 2011.
24. Green BM, Miles JC, Bradley EJ, Rees DM.
Radon Atlas of England and Wales. NRPB-W26. Chilton, United Kingdom: National Radiological Protection Board; 2002.
25. Miles JC, Howarth CB.
Validation Scheme for Laboratories Making Measurements of Radon in Dwellings: 2000 Revision. NRPB-M1140. Chilton, United Kingdom: National Radiological Protection Board; 2000.
26. Met Office. UKCP09: Gridded observation data sets. Available at:
http://www.metoffice.gov.uk/climatechange/science/monitoring/ukcp09. Accessed 24 August 2011.
27. Perry M, Hollis D. The generation of monthly gridded datasets for a range of climatic variables over the UK. Int J Climatol. 2005;25:1041–1054.
28. Bielsa I, Soria X, Esteve M, Ferrandiz C. Population-based incidence of basal cell carcinoma in a Spanish Mediterranean area. Br J Dermatol. 2009;161:1341–1346.
29. Health Protection Agency. UKRadon: Radon Action Level and Target Level. Available at:
http://www.ukradon.org/article.php?key=actionlevel. Accessed 24 August 2011.
30. Whitton JT. New values for epidermal thickness and their importance. Health Phys. 1973;24:1–8.
31. Perez-Losada J, Balmain A. Stem-cell hierarchy in skin cancer. Nat Rev Cancer. 2003;3:434–443.
32. Youssef KK, Van Keymeulen A, Lapouge G. Identification of the cell lineage at the origin of basal cell carcinoma. Nat Cell Biol. 2010;12:299–305.
33. Li JX, Chen R, Fu CL, Nie JH, Tong J. Screening of differential expressive genes in murine cells following radon exposure. J Toxicol Environ Health A. 2010;73:499–506.
34. Radespiel-Troger M, Meyer M, Pfahlberg A, Lausen B, Uter W, Gefeller O. Outdoor work and skin cancer incidence: a registry-based study in Bavaria. Int Arch Occup Environ Health. 2009;82:357–363.
35. Beral V, Robinson N. The relationship of malignant melanoma, basal and squamous skin cancers to indoor and outdoor work. Br J Cancer. 1981;44:886–891.
36. Lee JA, Strickland D. Malignant melanoma: social status and outdoor work. Br J Cancer. 1980;41:757–763.
37. Doherty VR, Brewster DH, Jensen S, Gorman D. Trends in skin cancer incidence by socioeconomic position in Scotland, 1978–2004. Br J Cancer. 2010;102:1661–1664.
38. van Netten C, Brands RB, Morley DR, Sabels BE. Development of a long-term personal radon monitor. Am Ind Hyg Assoc J. 1995;56:1107–1110.
39. Harley NH, Chittaporn P, Roman MH, Sylvester J. Personal and home 222Rn and gamma-ray exposure measured in 52 dwellings. Health Phys. 1991;61:737–744.
