Central nervous system (CNS) tumors are infrequent. Standardized incidence is 5.9 cases per 100,000 inhabitants in more developed countries for males and 4.4 for females, respectively, according to Globocan 2012.1 Mortality in developed countries is 4.0 and 2.7 per 100,000 inhabitants for males and females, respectively. The incidence rate in Europe for CNS tumors is 28.1 cases per 100,000 inhabitants aged 0-14 (adjusted to the world population).1 CNS tumors can be divided in different histology such as gliomas, neuronal tumors, poorly differentiated tumors, meningiomas and other tumors.

There is very scarce information on the risk factors for these tumors. Its low incidence makes difficult to have more information on their environmental and genetic factors. Some hereditary syndromes increase the risk such as neurofibromatosis, tuberous sclerosis, von Hippel-Lindau syndrome or retinoblastoma.2,3

According to the International Agency for Research on Cancer (IARC), there exists sufficient evidence to classify X and gamma rays exposure as risk factors for CNS tumors.4 This evidence comes mainly from studies based on medical treatment effects. The evidence is very limited with the use of mobile phones.5 Recent studies have observed that there is an association between the use of computerized tomography and brain tumors.6 Other authors suggest that vinyl chloride exposure could increase the risk of CNS tumors.7

Radon 222 is a chemical element pertaining to the noble gases family. Its short-life products Polonium 214 and Polonium 218 release radioactive alpha particles when they are transformed in other products. Radon 222 is the most frequent isotope of radon, comprising 80% of all radon isotopes and it is the most relevant from an epidemiologic point of view. Radon 222 is an odorless, colorless and tasteless gas that is present in the disintegration chain of Uranium 238. Uranium 238 is present in the earth crust rocks and indoor radon concentration depends mainly on the Uranium content of the rocks where a house has been built. Radon is denser than air and therefore radon concentration is usually higher in lower compared to upper floors. Alpha radiation has a poor penetrating capacity and produces damage to the lung epithelium when it impacts on it. There is an established association between radon exposure and lung cancer.8,9 Radon was declared a human carcinogen in 1987 and 1988 by the Environmental Protection Agency10 and by the IARC,4 respectively. It has been included as a risk factor to avoid in new version of the European Code Against Cancer released in October 2014.11 The WHO and EPA state that radon is the first risk factor of lung cancer in never smokers and the second in ever smokers.12 The presence of indoor radon varies depending on the uranium content and, in Spain, it is more frequent in Galicia, certain areas of Castilla y León, Northern Extremadura and Northern of the Comunidad de Madrid.13

Other organs different than the lung, including the kidney and the bone marrow, may receive low doses from radon exposure, as stated in page 14 of the WHO report on indoor radon.12 There are very few studies assessing the relationship between radon and CNS tumors. Radiation dose received by the brain due to radon and its byproducts is much lower than that received by other organs.14 Though definitions of low dose exposure are variable, we could define it as 10mSv, since this is equivalent to one computed tomography in one year. After one year of an average exposure of 200Bq/m3 (Becquerels per cubic meter) by inhalation, the brain would receive between 0.06 and 0.15 mSv whereas the lung in the same situation would receive between 35.8 and 159 mSv. Though there is not a clear pathogenic mechanism, it has been proposed that macrophages might phagocyte small solid particles in the lungs from radon descendants that could reach the CNS through the blood.14 Since ionizing radiations (and radon is an ionizing radiation) are a risk factor for CNS tumors, it is biologically plausible that radon could induce these tumors. Studies analyzing this possible association have shown discrepant results and have been performed mainly in miners and general population.

The aim of this study is to systematically review the existing evidence on the relationship between occupational and residential radon exposure (in adulthood or childhood) with central nervous system tumors.

We systematically reviewed the scientific literature. We used Medline (PubMed) and EMBASE for this purpose. We used various combinations of MeSH (Medical Subject Heading) terms combined with free text. Below appears the main bibliographic search used in Medline: ((“Central Nervous System Neoplasms”[Mesh] AND “Brain Neoplasms”[Mesh]) AND “Radon”[Mesh] OR “Radon Daughters”[Mesh]) AND (“humans”[MeSH Terms] AND (English[lang] OR Spanish[lang])).

We completed the search through consulting manually the references of the papers selected to be full-text read. The search period comprised from the first registries in both databases until 30/06/2016. A last update was performed in 21/11/2016. We put especial emphases in the exhaustiveness of the search in order to include all relevant information at the cost of obtaining not relevant information that had to be disregarded. In this case, the manual search of the references retrieved was especially relevant because most of the available papers did not have “central nervous system tumors” or “brain tumors” neither in the title nor in the abstract (or even brain tumors as a MeSH term). Therefore, we decided to look for all miners’ studies analyzing radon and different cancer types to check if in those studies CNS tumors had been also analyzed. A similar complementary approach was also used to locate papers for residential radon and CNS tumors in adults and in children. The potential papers to be included were checked by two reviewers, who decided by consensus if they should be included or not applying certain inclusion and exclusion criteria.

To our knowledge, this is the first systematic review assessing the possible relationship between radon exposure and CNS tumors. No clear association is apparent between radon exposure and the onset of CNS tumors in different settings. This unclear effect is observed when analyzing specifically the results of studies performed in miners, adult general population and children. The methodology of the available studies is highly heterogeneous and also the sample size and the number of cases included in the available studies.

Residential radon has been classified as a lung carcinogen and the scientific evidence is scarce regarding a possible effect on other cancers. Some studies have pointed to a possible association with leukemia,35 skin cancer36 or esophageal cancer.37 The most difficult issue when assessing a possible association with a cancer different than lung cancer is that the effective doses reaching other organs are much lower than that received by the lungs. The studies performed in miners have the advantage that radon exposure is usually higher than that found in residential settings and this aspect makes easier to establish dose-response patterns. Nevertheless, miners’ studies also face the disadvantage that participants are also exposed to other carcinogens such as silica dust, gamma radiation (uranium miners) and other dusts that can be found in mines. All these studies have been performed in males and the number of cases found in these cohorts has been very low in many of them, leading to imprecise estimations. Radon exposure assessment has been variable in these studies. Some of them have used estimations of exposure,18,19 while others have measured individual exposure23 or measured exposure for a part of the follow-up duration.17,20–22,24 Other possible biases might be present in miners’ studies regarding tumor classification. The study by Tirmarche et al.17 used International Classification of Diseases (ICD), 8th version, (that included jointly tumors of unknown nature of the eye, brain or other parts of the CNS). Other studies20,24 employed ICD-9th version (neoplasms of undetermined nature of the brain and benign neoplasms of the CNS). Other studies used other classifications such as ICD-10th version or included only malignant tumors. The investigation by Schubauer-Berigan et al.,23 which observes a negative association between radon and CNS tumors, used the Minor Cause of Death from the National Institute for Occupational Safety and Health Life that includes tumors from the peripheric nervous system. The study by Darby et al.18 did not use the ICD-9th version. These differences difficult the comparison of the results found in miners.

Finally, none of the studies performed in miners had as the main hypothesis to analyze the possible relationship between radon exposure and CNS tumors. Since these tumors have a very low incidence, a cohort study is not probably the best design. A case-control study is a more efficient option. The number of observed deaths is also very low making difficult to obtain any association in these cohorts. If we compare the results observed for overall cancer risk in miners’ studies with the specific result observed for CNS risk, 3 studies17,18,20 have observed a higher risk for CNS tumors while the remaining observed a similar but not a lower risk. It seems therefore that the risk for CNS tumors for radon exposure in miners might be slightly higher than that observed for overall cancer risk in this population. We do not believe that publication bias is present in the relationship between radon and CNS tumors, mainly because this association has been sparsely studied and a specifically designed investigation on this issue should be welcomed by most journals.

There were only three studies assessing residential exposure with CNS tumors in adults. One study measured radon in water and was published decades ago while the other two are very recent and have discrepant results, with no association and a significant dose-response effect. Both studies have a good methodology and the main difference between them is how radon exposure was measured. While in the study by Turner et al.,26 which included a high sample size, mean county radon exposure was assigned to each participant, in the study by Bräuner et al.27 residential radon exposure was assigned individually to each participant after applying a validated model used to estimate radon exposure taking into account different variables.38 The possibility of non-differential exposure misclassification cannot be also excluded when assigning individual residential radon exposure. Another important difference is that the study by Turner et al.26 included participants that were friends or relatives of the selected volunteers, and therefore a potential selection bias could be present, reinforced by the fact that 30% of participants of the cohort were college graduates, and therefore might not represent properly US adult population. Again, there are differences in how the different CNS tumors have been considered as cases, since the included studies have used different classifications. Finally, the Bräuner et al.27 study considered more adjustment variables than the Turner et al.26 study, including diet, occupation, education and exposure to nitrogen oxides.

The studies performed in pediatric population have discrepant results. A proof of the interest in this possible association is the number of available studies, eight, compared with three performed in adults. The main difference with CNS tumors in children vs adults is that the most frequent tumor is meduloblastoma instead of astrocytoma.39 Children are also more sensitive than adults to ionizing radiation because they have a higher number of cells in division compared with adults. An important issue is that the levels of indoor exposure were highly variable in the included studies, i.e. the study by Kendall et al.32 had an average radon exposure of 21Bq/m3 while the study by Hauri et al.28 had an average of 77, close to four times higher. It is important to mention that all studies have been performed in areas that have not been classified as radon prone areas and therefore studies performed in radon prone areas might pose different results. There was also high variability on the design and sample size of the different studies. The investigations with the highest quality were the Swiss and Norwegian cohorts and two case-control studies performed in Denmark and England. None of these investigations directly measured radon exposure in the children's dwellings and radon exposure was estimated using different models. The prediction model for indoor radon exposure in the Hauri et al.28 study had a sensitivity of 0.29-0.31 and a specificity of 0.92. The prediction model by Raaschou-Nielsen et al.33 used in the Danish study correctly predicted 80% of high exposures and 60% of low radon exposures. Of note, the study by Kaletsch et al.,30 which measured radon exposure individually observed a statistically significant odds ratio of CNS tumors of 3.85, being this the highest in all studies. Nevertheless, the study by Cartwright et al.,31 which also measured radon exposure individually, did not find any association. The possibility of a selection bias in the studies by Raaschou-Nielsen et al.,33 Hauri et al.28 and Kendall et al.32 is not present because they used information from the census or from a tumor registry.

The great majority of the included studies have adjusted their results for many covariates. Miners’ studies have adjusted by age, ethnicity, tobacco consumption and gamma rays exposure whereas studies in adult population have adjusted their results by age, gender, tobacco consumption, education and socioeconomic status. Finally, studies performed in children have also adjusted their results by many covariates. These adjustments probably mean that the results observed are not probably due to unmeasured confounding factors and support a low risk of bias in the included studies. This risk is higher in miners’ studies compared to studies performed in adults and in children, a fact reflected in the scoring of each category studies. Most of them did not adjust their results by tobacco consumption or had this information for a limited number of participants. One added limitation is the low incidence of these tumors which produces imprecise estimations in most studies.

The quality of the included studies has been highly variable, with a score between 4 and 9 points. The scoring is particularly heterogeneous for miners and studies performed in children, with studies performed in adults of the general population having a similar quality. Though there are some aspects that we have considered in our quality scale that could be discussed, in our opinion they reflect the main aspects of the included studies such as sample size, number of cases included, study design and number of covariates considered.

The present systematic review has some advantages. To our knowledge, it is the first study reviewing the existing evidence on radon exposure and CNS tumors. We have also developed a quality scale to assess the differences between the studies included and also to compare is validity. The classification of the studies retrieved in occupational, adults and pediatric population, allow us to consider the particular characteristics of these individuals.

This study has some limitations. The main one is that it has not been possible to meta-analyze the included studies, neither globally nor by the subgroups included. This is due to the high heterogeneity of the studies, which are case-control studies, cohort studies or ecological studies. Besides the individual limitations of each study, there is the possibility that we have missed studies published in languages different than English or Spanish. Nevertheless, we are not aware of any study that we have excluded due to language limitations, since we should have found this study in the literature search. A further limitation is that many studies have adjusted their results by few covariates (age and gender mainly) and some of them did not adjust their results for tobacco consumption.

There are few studies that have assessed the relationship between radon exposure and CNS tumors and the results are contradictory. Some rigorous studies have observed statistically significant associations in miners or in general population. Nevertheless, other high quality studies have not observed any association. Two of the studies that found an association were performed in women and children, populations that are not available in miners’ studies. Residential studies are limited because it is difficult to estimate radon exposure. Other problem is that CNS tumors comprise many different diseases and the tumors included are different in some studies.

We can conclude that it is necessary to perform more robust studies, preferably in areas with high radon concentrations and using preferably individual measurements to assess radon exposure. Given the low incidence of CNS tumors and the existence of biological plausibility for a potential effect, more research is needed on this topic.

Maria-Victoria Zunzunegui.

All authors have contributed equally to the paper's content.

Part of this work has been performed during a Fulbright grant for Senior Researchers awarded to Alberto Ruano-Ravina at Brown University (Providence, Rhode Island, USA) (REF PRX14/00365). This grant is awarded by the US Department of State in collaboration with the Spanish Ministry of Education.

None.