A restatement of the natural science evidence base concerning the health effects of low-level ionizing radiation

Exposure to ionizing radiation is ubiquitous, and it is well established that moderate and high doses cause ill-health and can be lethal. The health effects of low doses or low dose-rates of ionizing radiation are not so clear. This paper describes a project which sets out to summarize, as a restatement, the natural science evidence base concerning the human health effects of exposure to low-level ionizing radiation. A novel feature, compared to other reviews, is that a series of statements are listed and categorized according to the nature and strength of the evidence that underpins them. The purpose of this restatement is to provide a concise entrée into this vibrant field, pointing the interested reader deeper into the literature when more detail is needed. It is not our purpose to reach conclusions on whether the legal limits on radiation exposures are too high, too low or just right. Our aim is to provide an introduction so that non-specialist individuals in this area (be they policy-makers, disputers of policy, health professionals or students) have a straightforward place to start. The summary restatement of the evidence and an extensively annotated bibliography are provided as appendices in the electronic supplementary material.

with the accompanying editorial Wakeford (2015). Doll (2002) gives a further review. 3. UNSCEAR (2008) gives global figures on natural and artificial background radiation. Abbot (2015) gives comparative figures for different countries and different years. In this paper we define, for sparsely ionizing radiation, a low dose as being <100 mGy and a low dose-rate as being <0.1 mGy/min averaged over one hour, following UNSCEAR (2012b). of alternative theories about the dangers (or otherwise) of radiation do not view reports by these international bodies as authoritative and instead view them as representing an establishment position that attempts to argue 'from authority'. However, these reports are widely recognised as representing the views of the majority working in the field. 9.
For an example of an optimization procedure in the context of management of occupational exposure, see Figure 1 in IAEA (2002). f.
For example, the average annual dose to occupationally exposed workers in the UK is 0.0004 mSv Oatway et al. (2016). For sources that arise from the disposal of radioactive waste the recommended effective dose constraint for public exposure is 0.3 mSv/yr and for prolonged exposure from long-lived radionuclides, if dose assessment is problematic, the recommended dose constraint is <0.1 mSv/yr. A radionuclide is a particular version of an atomic nucleus characterised by its number of protons and neutrons and their arrangement within the nucleus. For example iodine-131 ( 131 I) is the radionuclide of iodine with 53 protons and 78 neutrons, whereas the nucleus of technetium-99m has 43 protons and 56 neutrons in a metastable state. g.
Lecomte et al. (2014) refers to an upper reference level of 300 Bq/m 3 for radon-222 corresponding to about 4.5 mSv/y in workplaces and 15.8 mSv/y in homes. Radon-related risk in the home is determined by a person's smoking status as well as indoor radon concentration Gray et al. (2009). The phrase "according to the situation" is a quote from Table 8 ICRP 103 which refers to the judgement required when levels of radiation are abnormally high. h. Chapter 7 of ICRP 103 (2007) describes the system of recommendations for medical exposures. See also Wrixon (2008). 10. Brenner (2010), Muller (2015), Auvinen et al. (2015). 11. Morgan and Bair (2013), Niles (2014), Pearce (2015). Annex IV of UNSCEAR (2008) and the WHO (2006) Chernobyl Forum report both document the large health impact of the Chernobyl accident because of fears about radiation. 12. Authors' summary.  ) found a linear doseresponse relationship between radiation-related risks and major coronary events, suggesting that the LNT concept may be relevant to some forms of circulatory disease, although this was a study of high doses received during radiotherapy.
19. See UNSCEAR (2006a) pages 32-33 on human genetic susceptibility. Susceptibility and the development of biomarkers to identify radio-sensitive cancer patients is discussed in Manning and Rothkamm (2013). a. Whilst the excess relative risk of cancer is higher for exposures in childhood than for those in adulthood, the difference in excess absolute risk is not as marked (see paragraph 37b). Ozasa (2015). Kashcheev et al. (2015) describe risks in Chernobyl workers who cleaned up after the accident. Mayak workers at the eponymous nuclear plant were subject to prolonged low dose-rate external gamma radiation and plutonium exposure Sokolnikov et al. (2015), as were nearby Techa River residents due to discharges of radioactive waste Davis et al. (2015).
Yangjiang is an area of high natural background radiation in China Tao et al. (2012) and Kerala is an area of high natural background radiation in India Nair et al. (2009). Ankylosing spondylitis patients in the UK were historically treated with X-rays in the mid-20 th century Weiss et al. (1994). For the purposes of representation and as the studies pertain almost entirely to low-LET radiation, Sv and Gy are assumed to be equivalent in Figures 2a and 2b. For the purposes of representation, ERRs for cancer incidence and mortality are also plotted on the same axis. In reality, the relationship between cancer incidence and mortality will depend on the ability and availability of diagnostics and therapies to improve survivorship. See Coleman et al. (1993). For underground miners, risks are expressed relative to cumulative radon exposures given in WLM, and for domestic exposures to radon, risks are expressed relative to concentrations in Bq/m 3 . These diverse studies include individuals who have been exposed at greatly varying dose-rates -some briefly, others very slowly. Ruhm et al. (2016) summarise recent evidence and debate on the impact of dose-rate upon the health effects of radiation, although a distinction needs to be made between low-LET and high-LET radiation. Table 8 contains further detail about the data in this figure.        Table 1 are primarily of concern in partial body irradiation, for instance erythema (skin) and permanent sterility (gonads). Radiation effects upon cataracts and circulatory disease do not fall neatly into either tissue reaction or stochastic effects. 29. Streffer et al. (2003). Chapter 8 in Mettler and Upton (2008) gives a detailed review of tissue reactions of in utero exposure to radiation. Otake and Schull (1998) review radiation-related brain damage and growth retardation amongst prenatally exposed survivors. 30. Otake et al. (1990) describe studies of untoward pregnancy outcome (defined as stillbirth, major malformation or death within 14 days) amongst >65,000 offspring of atomic bomb survivors finding a positive association with joint parental dose, but the regression slope was not statistically significant.  Table 9. After the bombs were detonated 60-80,000 people were killed instantly in Hiroshima and another 90-166,000 died in the ensuing 4 months. In Nagasaki there were 22-75,000 instant fatalities and another 60-80,000 deaths in the ensuing months. The LD50 dose (at which 50% of the exposed population died) occurred at a radius of 1-1.3km of each blast and later dose reconstruction yielded a bone marrow dose estimate for the LD50 of 2.9-3.3 Gy Pierce et al. (1996), Preston et al. (2003), Wakeford (2004). The key message from the LSS papers is that risk of cancer mortality for people exposed on the day remains elevated 60 years on. individuals situated within 2.5 km of the blast were exposed to levels of radiation of 5 mGy or higher with a mean dose of 200 mGy. The 38,500 people 2.5-10 km from the blast received doses below 5 mGy. The 26,500 people not in the city were unexposed residents of Hiroshima or Nagasaki who were not in either city ('NIC') at the time of the bombings. 85% of the cohort experienced irradiation below the mean level of 200 mGy. Dose distribution of the cohort is given in Table 1 Table 9. for in utero exposure. The EAR did not increase with time/age as it did for an equivalent cohort exposed in early childhood, suggesting that lifetime risks following in utero exposure may be lower than for early childhood exposure Ohtaki et al.   Authors' summary. The excess relative risk quoted here is different from the nominal cancer risk coefficient of 5.5% per Sv derived by the ICRP and used in optimization calculations. The ERR of 0.47 per gray for solid cancer is an estimate of the amount by which the underlying risk of solid cancer is increased proportionally for each gray of exposure. The ICRP's "nominal risk coefficient" of 5.5% per Sv is an estimate of the health detriment due to cancer experienced in a population exposed to low level radiation; it includes attributable fatal and non-fatal cancer, years of life lost and pain and suffering.  report that ERR/Gy is decreasing with time since exposure in a Russian cohort exposed as children.

UNSCEAR (2008)'s
Suzuki and Yamashita (2012) summarises the debate about low dose risk of thyroid cancer, concluding that a statistically significant increase has hardly been described with radiation doses below 100 mSv. A recent pooled analysis of 12 studies of thyroid cancer after childhood exposure to external radiation found a significant increase in RR for doses <0.10 Gy with no significant departure from linearity Veiga (2015) and a longer one in part II (page 28) of UNSCEAR (2013a). 72. Table 2 of Hasegawa et al. (2015), summarised in Table 12 below.  The large pooled study, INWORKS, includes studies of worker cohorts in the UK, France and the US. The results for the nuclear worker studies and the LSS are in broad agreement even though the workers usually accumulated their doses over many years, whilst the LSS subjects received theirs in a few seconds -supporting the assumption of additivity and dose-rate independence of radiation doses. There are other key differences between the LSS and worker studies, including exposure to different types of radiation (e.g. gamma rays of different energies), and differences in the demographics, genetics and lifestyle features of the subject population Stewart and Kneale (2000), Little (2002a). The LSS figures for working-age males in Table 4   See Table 3 in Shore (2016b) for a summary that suggests that there is occupational radiation cataract risk amongst medical specialists who receive large cumulative doses (with estimated mean doses from various studies ranging from 0.028 Gy to 6 Gy). 84. 10 Gy is the average alpha dose to the skeleton as a whole: to endosteal surfaces, the putative originating cells for osteosarcoma, the average alpha dose would be about half this. In a US cohort of 820 people there were 46 deaths from bone cancer where less than 1 would have been expected, and a clear excess of cancers of the paranasal sinuses and mastoid air cells was also apparent due to radon formed on the decay of 226    The results of these studies are not statistically significant, although they are of comparable size and dose to some of the worker studies that have identified positive, statistically significant estimates of ERR/Gy. Nevertheless, there is no statistical inconsistency between these estimates, although the Kerala risk estimate is close to statistical incompatibility with the LSS risk estimate. 90. A case-control study in Great Britain based on >27,000 cases from the National Registry of Childhood Tumours compared risks of childhood leukaemia and other cancers with cumulative dose through exposure to indoor gamma radiation and radon based on the mother's address at the time of the child's birth Kendall et al. (2013). It found a statistically significant relationship between dose from naturally occurring gamma radiation and the risk of childhood leukaemia, with an ERR/Sv = 120 (95% CI: 30 to 220). Radon exposure did not predict childhood leukaemia, and other childhood cancers were not related to either radon or gamma radiation exposure. UNSCEAR (2013b) cautions that there are large uncertainties associated with this study with respect to its use of an ecological measure of dose. A census-based cohort study of childhood cancer in Switzerland (with 1,800 incident cases) reported positive relationships between cumulative dose of external radiation and both childhood leukaemia (ERR/Sv = 50 (95% CI: 0 to 100)) and central nervous system cancers (ERR/Sv = 50 (95% CI: 0 to 110)) Spycher et al. (2011). A Finnish case-control study (1093 cases) with full residential history found a non-significant odds ratio increase for childhood leukaemia with increasing dose-rate of background radiation, with a significantly elevated odds ratio in the age group 2-7 years Nikkila et al. (2016). These relationships between exposure to background gamma radiation and childhood leukaemia incidence are broadly comparable to those from the LSS, lending some support to the application of risk estimates derived from the LSS to the very low doserates received from naturally occurring background gamma radiation. However, a French census-based analysis with 9,056 incident cases over 20 years found no evidence of an association of childhood leukaemia risk with either radon (SIR by 100 Bq/m 3 1.01, 95% CI 0.91 to 1.12) or gamma radiation (SIR by 10 nSv/h 1.01, 95% CI 1.00 to 1.02) Demoury et al.  (2013)). The geographical approach was also repeated for the same risk group in France, Britain and Switzerland. Neither case-control study recapitulated the odds ratio (OR) significantly different from 1 and just one time interval from the French geographical study generated an SIR marginally significantly different from 1. When the French study broadened the age group under consideration to include children under 15, a marginally significant result was observed (see Table 2

MEDICAL EXPOSURE
96. Dose fractionation (in which the total dose is delivered as a number of doses separated in time) allows the optimization of the lethal effect on diseased cells while sparing healthy tissues. There are known risks from such therapy which have to be balanced against the benefits of treating the underlying disease. There is a large body of data on those risks which is growing as radiotherapy becomes more successful and people survive ever longer after their radiotherapy. These data have to be treated with caution as individuals treated with radiotherapy are already patients, so they are not a representative sample of the general population, and this could affect estimates of radiation risks. Further, radiotherapy is usually focused on localised diseased tissues, leading to a highly heterogeneous distribution of doses within the body. There is an overall pattern that the ERR/Gy from radiotherapy tends to be lower than the corresponding values in the LSS. This pattern is more marked at higher average radiotherapy dose and is therefore thought to be explained by spatially-focussed radiation used in therapy killing a large proportion of cells that might otherwise have become cancerous due to irradiation -the so called sterilization effect.
However these two patterns are not ubiquitous: some individual radiotherapy studies have a higher ERR/Gy than the corresponding values in the LSS. Little (2001) reviews 116 radiotherapy studies and compares ERRs for incidence and mortality with comparable risks in the LSS. Travis et al. (2003a) found that radiation related risk remained high even at the highest doses in women < 30 years of age treated for Hodgkin disease with radiotherapy (i.e., there was no evidence of a sterilization effect). Wakeford (2004) Table 4 gives estimates for dose from various different examinations. a. Bithell and Stewart (1975) (2015). A further overview is Boice (2015). 99. Authors' summary. conducted a meta-analysis of low doserate epidemiologic studies that provide dose-response estimates of total solid cancer risk in adulthood in comparison to corresponding acutely-exposed atomic bomb survivor risk, in order to estimate a dose rate effectiveness factor (DREF) of between 1 and 2. Ruhm et al. (2015) describe the historical development of the DDREF concept in light of emerging scientific evidence on dose and dose-rate effects, summarises the conclusions recently drawn by a number of international organisations, mentions current scientific efforts to obtain more data on low dose and low dose-rate effect effects at molecular, cellular, animal and human levels, and discusses future options to improve and optimize the DDREF concept for the purpose of radiological protection. 123. Authors' summary. 2. Abylkassimova Z., Gusev B., Grosche B., Bauer S., Kreuzer M., Trott K. 2000 Nested case-control study of leukemia among a cohort of persons exposed to ionizing radiation from nuclear weapon tests in Kazakhstan (1949Kazakhstan ( -1963. Ann Epidemiol 10, 479.