Human Risk Assessment Radiation Epidemiology

Transcription

1 Human Risk Assessment Radiation Epidemiology Richard Wakeford Professor in Epidemiology, Institute of Population Health and Dalton Nuclear Institute, The University of Manchester, UK

2 Scientific Assessment and Radiological Protection The evidence for the risk to human health from exposure to ionising radiation is reviewed regularly by expert committees. The conclusions of these scientific reviews are used by radiological protection groups principally the International Committee on Radiological Protection (ICRP) to produce appropriate schemes for protecting workers and members of the general public from the harmful effects of radiation.

3 Scientific Assessment United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) issues reports a few times a decade Other international assessments (e.g. International Agency for Research on Cancer (IARC))

4 Scientific Assessment US National Academy of Sciences Committee on the Biological Effects of Ionizing Radiations (BEIR) dels.nas.edu/dels/rpt_briefs/beir_vii_final.pdf US National Council on Radiation Protection and Measurements (NCRP)

5 Scientific Assessment Centre for Radiation, Chemical and Environmental Hazards of Public Health England (PHE) (formerly HPA, formally NRPB) French Académie des Sciences and Académie Nationale de Médecine

6 Radiological Protection International Commission on Radiological Protection (ICRP) Public Health England US National Council on Radiation Protection and Measurements (NCRP)

7 Radiation Effects Adverse radiation-induced health effects fall into two classes: Deterministic effects Stochastic effects

8 Radiological Protection The principal objectives of radiological protection are to Prevent deterministic effects Limit stochastic effects

9 Radiation Dose The absorbed dose (D) of ionising radiation is the energy deposited per unit mass of material: 1 gray (Gy) = 1 J/kg. The equivalent dose (H) is the absorbed dose adjusted by the radiation weighting factor (w R ), which takes account of the biological damage relevant to stochastic health effects caused by particular radiations (e.g. the w R for sparsely ionising γ-rays is 1, but the w R for densely ionising α-particles is 20). The equivalent dose is measured in sievert (Sv).

10 Radiation Dose The effective dose (E) is the sum of organ/tissue equivalent doses, each weighted by the tissue weighting factor (w T ), which take account of different tissue sensitivities. SI unit for both H and E is the sievert (Sv). H and E are for stochastic effects in humans only. H and E are radiological protection quantities that should be used outside the ICRP system with great caution.

11 Deterministic Effects Severe impairment of function of organ or tissue Threshold dose below which the effect does not occur Severity of effect increases with dose Cell killing is the underlying mechanism

12 Deterministic Effects Now also referred to as early or late harmful tissue or organ reactions Erythema is a typical early reaction Deep tissue necrosis is a typical late reaction

13 Deterministic Effects Thresholds (ICRP Publication 118, 2012) Sterility permanent acute dose 3-6 Gy temporary (testes) acute dose 0.1 Gy Lens of the Eye cataract acute dose 0.5 Gy Bone Marrow depression of haematopoiesis acute dose 0.5 Gy Skin burns acute dose 5-10 Gy

14 Intrauterine Irradiation Irradiation of the conceptus, embryo and fetus requires the consideration of particular effects related to the development of the organism. A reduced lethal dose, developmental malformations and mental retardation need to be considered in addition to other deterministic and stochastic effects.

15 Deterministic Effects Death after whole-body exposure half the exposed individuals in a healthy adult human population would be expected to die within 60 days of receiving an acute dose of 4 Gy due to bone marrow damage, i.e. the LD 50/60 is ~4 Gy.

16 Mortality Whole body absorbed dose, Gy Principal effect contributing to death Time to death, days 3-5 Damage to bone marrow Damage to GI tract Damage to lungs and kidneys >15 Damage to nervous system <5, dose-dependent NMBU, Ås, Norway, 15 June 2017

17 Radiation Accidents (I. Turai et al., BMJ 2004; 328: ) Since 1944 there have been 430 serious accidents involving radiation. These accidents have involved about 3000 persons and have resulted in 134 deaths. Immediate problem is deterministic effects, but risk of stochastic effects is also raised. See also: Nénot JC, Radiation accidents over the last 60 years. J Radiol Prot 2009; 29:

18 Radiation Accidents Nuclear energy or weapons accidents reactor accidents (e.g. Chornobyl, Fukushima) criticality accidents (e.g. Tokai Mura) Industrial, mainly NDT (about half of all events) Medical uses account for about one-tenth of events, but erroneous calibration and insecure sources are the cause of nearly half of all deaths.

19 Stochastic Effects Cancer and hereditary abnormalities No threshold dose Probability (but not severity) of effect increases with dose Non-lethal cell modification (mutation)

20 Stochastic Effects Laboratory experiments (in vivo and in vitro) have demonstrated unambiguously that ionising radiation can produce stochastic effects IARC Group 1 carcinogen ( carcinogenic to humans ). However, it is difficult to predict the level of these effects in humans purely from data based upon experimental studies.

21 Epidemiology Epidemiology is the study of the distribution of disease in human populations and of the factors that determine the risk of disease in these populations. Epidemiology considers groups of humans rather than individuals. Epidemiology is largely an observational (i.e. non-experimental) science.

22 Observational Science Observational sciences, unlike the experimental sciences (e.g. randomised controlled clinical trials), have to use data generated by the uncontrolled conditions of everyday life. Randomisation is not possible in the observational sciences. This makes the reliable interpretation of epidemiological results more difficult because the insidious effect of extraneous factors is more uncertain.

23 Interpretation of Results A statistical association found in an epidemiology study does not necessarily imply an underlying direct cause-and-effect relationship. The association could be due to Chance Bias Confounding (an indirect causal effect) Some combination of these

24 Chance Tests of statistical significance provide a measure of the probability that an association between the rate of disease and a certain factor has been produced by chance alone and that there is no underlying relationship between the risk of the disease and the factor (e.g. a particular exposure).

25 Bias A bias (or systematic error) in a study leads to an inaccurate result (at some level) because of the input of incorrect information. This can arise in a number of ways, such as better disease ascertainment in an exposed group than in an unexposed group, or better exposure information for affected individuals than for unaffected individuals. The elimination of sources of bias is an important aspect of study design.

26 Confounding Confounding occurs when a statistical association between a disease and a certain factor is actually produced by a related factor. As a hypothetical example, sclerosis of the liver might be found to be associated with the drinking of tonic water, but is actually caused by the alcohol in the gin or vodka that is drunk with the tonic water.

27 Confounding Factors A statistical association may be adjusted (to some degree) by controlling for the presence of known potential confounding factors (e.g. smoking) in the statistical analysis of the factor under study. To do this, appropriate data must be available for the known potential confounding factors. Clearly, unknown confounding factors cannot be dealt with in this way.

28 Sir Austin Bradford Hill s Causal Guidelines (A B Hill, The environment and disease: association or causation? Proc Roy Soc Med 1965; 58: ) Temporality Biological plausibility Strength of association Consistency Dose-response Coherence Specificity Analogy Human experiments

29 Bradford Hill s Causal Guidelines Temporality The effect must follow the cause after an appropriate period (e.g. an appropriate latent period for a cancer) Strength of Association A stronger statistical association is more likely to be causal Consistency Is the association found repeatedly under different circumstances?

30 Bradford Hill s Causal Guidelines Dose-response ( biological gradient ) Does the effect increase as the exposure is increased? Biological plausibility A causal interpretation is assisted by a recognised biological mechanism to produce the effect Coherence The association should not conflict with established understanding of the disease

31 Bradford Hill s Causal Guidelines Specificity The association is specific to a particular disease or set of diseases rather than a range of diverse diseases Analogy Is the association similar to an established cause-and-effect linkage? Human experiment A material change in the exposure should lead to the anticipated change in the disease rate

32 Bradford Hill s Causal Guidelines None of my nine viewpoints can bring indisputable evidence for or against the causeand-effect hypothesis and none can be required as a sine qua non. What they can do, with greater or less strength, is to help us to make up our minds on the fundamental question is there any other way of explaining the set of facts before us, is there any other answer equally, or more, likely than cause and effect? Sir Austin Bradford Hill (1965) A B Hill, The environment and disease: association or causation? Proc Roy Soc Med 1965; 58:

33 Tests of Statistical Significance is there any other way of explaining the set of facts before us, is there any other answer equally, or more, likely than cause and effect? No formal tests of significance can answer those questions. Such tests can, and should, remind us of the effects that the play of chance can create, and they will instruct us in the likely magnitude of those effects. Beyond that they contribute nothing to the proof of our hypothesis. Sir Austin Bradford Hill (1965) A B Hill, The environment and disease: association or causation? Proc Roy Soc Med 1965; 58:

34 Epidemiology Epidemiological results are derived directly from the study of groups of humans, the species of interest. Epidemiological findings do not require the generalisation of results of experimental studies (e.g. of laboratory animals) to humans.

35 Radiation Epidemiology At present, a radiation-induced stochastic health effect cannot be distinguished from the same effect produced by some other factor. Cancer risks must be determined from the study of suitably exposed human populations.

36 UNSCEAR 1994 Report Studies of disease in human populations must adhere strictly to epidemiological principles in order to achieve valid quantitative results. These include sound case ascertainment, an appropriate comparison group, sufficient follow-up, an accounting for confounding factors and well-characterised dosimetry.

37 Descriptive Epidemiology In descriptive epidemiology, data are collected and collated and then examined for patterns of interest, such as variation of rates with time, place or race. From these patterns, hypotheses may be formulated that can be tested through suitably designed epidemiological studies.

38 Cancer Incidence in the UK Number of new cancer cases diagnosed and age-specific incidence rates per 100,000 population. All cancers excluding non-melanoma skin cancers (NMSC), by sex, in the UK in , male cases Number of new cancer cases 20,000 15,000 10,000 5,000 female cases male rate female rate Rate per 100,000 population Age at diagnosis (years)

39 Cancer Incidence in the UK Breakdown (%) of All New Incident Cases of Cancer Diagnosed in the UK during 2000 by Sex and Type. Lung Colorectal Prostate Breast Bladder Ovary Stomach Uterus Head and Neck Non-Hodgkin's Lymphoma Oesophagus Leukaemia Kidney Pancreas Melanoma Cervix Brain and CNS Testis Multiple Myeloma Mesothelioma Liver Hodgkin's Lymphoma Thyroid Males Females Percentage of All Cancers (except NMSC)

40 Cancer Mortality in USA

41 Age-specific mortality rates of lung cancer, males, England and Wales, Rate per 100,000 males and over Year of death

42 Age-specific mortality rates of lung cancer, females, England and Wales, and over Rate per 100,000 population Year of death

43 Childhood Leukaemia Rate of Leukaemia Mortality among Children 0-14 Years of Age in England and Wales during the Twentieth Century Mortality ( ) 30 Rate per million person-years Calendar Year

44 Childhood Leukaemia Trends (Stewart and Kneale, Nature 1969; 223: ) Rate of Leukaemia Mortality by Year of Age among Children 0-7 Years of Age in England and Wales in Six Successive Periods during Rate of Leukaemia Mortality, deaths per million person-years Age at Death, years

45 Childhood Leukaemia by Race and Age Average Annual Rate of Incidence of Childhood Leukamia in the USA by Race and Age. Both Sexes. (SEER Data for and ) 130 Incidence Rate, cases per million person-years White Black Age (years) at Diagnosis

46 Childhood Leukaemia by Type and Age Average Annual Rate of Incidence of Childhood Leukamia in the USA by Type and Age. All Races, Both Sexes. (SEER Data for and ) 100 Incidence Rate, cases per million person-years ALL AML CML Age (years) at Diagnosis

47 Childhood ALL by Subtype (Greaves et al., Leukaemia Res 1985; 9: ) 175 Frequency Distribution of Acute Lymphoblastic Leukaemia Subtypes by Age at Diagnosis Number of Immunophenotyped Patients call T-ALL Null ALL Age at Diagnosis (years)

48 Acute Lymphoblastic Leukaemia Major molecular subsets of ALL in infants (<1 year old), children (2-10 years old) and adults (Greaves, BMJ 2002; 324: )

49 Childhood Leukaemia Rate of Leukaemia Mortality among Children 0-14 Years of Age in England and Wales during the Twentieth Century Mortality ( ) 30 Rate per million person-years Calendar Year

50 Childhood Leukaemia Rate of Leukaemia Mortality among Children 0-14 Years of Age in England and Wales during the Twentieth Century Mortality ( ) 30 Rate per million person-years Calendar Year

51 Childhood Leukaemia Rates of Leukaemia Mortality and Registered Incidence among Children 0-14 Years of Age in England and Wales during the Twentieth Century Mortality ( ) Incidence ( ) 30 Rate per million person-years Calendar Year

52

53 Group Correlation Studies Group correlation ( ecological ) studies examine the variation (e.g. geographical, temporal) in disease rates and in potential causal factors for groups of people with the objective of indentifying potential cause-and-effect relationships. The results of group correlation studies must be treated with caution.

54 Group Correlation Studies Group correlation studies make the assumption that group-averaged exposures apply to affected individuals. This may not be correct and could lead to misleading inferences. The findings of group correlation studies must always be confirmed by individualbased studies before being accepted as valid they are hypothesis-generating studies.

55 Age-specific mortality rates of lung cancer, males, England and Wales, Rate per 100,000 males and over Year of death

56 Age-specific mortality rates of lung cancer, females, England and Wales, and over Rate per 100,000 population Year of death

57 Annual Consumption of Manufactured Cigarettes per Person, UK,

58 Cancer Mortality in USA

59 γ-ray Exposure in USA

60 Lung Cancer Mortality in USA

61 Radon Exposure in USA

62 Radon and Lung Cancer (geographical correlation study, US counties) (Cohen, Health Phys 1995; 68: )

63 Study of Puskin (JS Puskin, Health Physics 2003; 84: ) Puskin applied the analysis structure of Cohen to cancers other than lung cancer. Found the same pattern as lung cancer for other smoking-related cancers, but not for cancers unrelated to smoking. These other smoking-related cancers are not found to be related to radon exposure in underground hard-rock miner studies.

64 Cohort and Case-control Studies Cohort and case-control studies are individual-based epidemiological studies that use information (e.g. exposures) for individuals rather than groups. These studies are hypothesis-testing studies (or analytical studies).

65 Cohort Studies A cohort ( follow-up ) study examines the disease rates within a defined group of individuals with known levels of exposure, e.g. the cancer rates among workers at a particular nuclear installation during a specified period with various individuallydetermined doses of radiation. Exposure levels are determined before the disease outcome is known. Carefully designed cohort studies provide some of the firmest epidemiological evidence.

66 Case-control Studies A case-control study starts from a defined set of cases of a particular disease and derives for these affected individuals exposures that are of potential relevance. A set of controls representative of the population from which the cases are drawn is then selected, and the exposures of these controls are derived.

67 Case-control Studies The frequency with which an exposure occurs among the cases is then compared with that among the controls, e.g. the level of exposure to radon among lung cancer cases as compared to that among controls. Causal factors will have a stronger representation among cases than among controls.

68 Cohort vs Case-control Studies Cohort studies are large and expensive, and often require a long period of followup before meaningful results are obtained. However, the establishment of a database of individual exposures before the disease outcome is known makes the avoidance of biases easier to achieve.

69 Cohort vs Case-control Studies A case-control study can concentrate upon particular individuals (the cases and the controls) making exposure assessments more detailed and rigorous. This is especially so for a rare disease. However, case-control studies can be more prone to biases, e.g. are the controls appropriate or is the exposure information as accurate for controls as for cases?

70 Cohort vs Case-control Studies A case-control study is sometimes nested within a cohort, usually to reduce the workload associated with a full cohort study. Such nested case-control studies are less prone to biases because the controls are generally more representative of the population from which the cases are drawn the controls may be a representative sample of the cohort but cases and controls should always be treated equivalently to avoid biases.

71 Cohort vs Case-control Studies The potential for biases and uncontrolled confounding to affect the results of both cohort and case-control studies must always be carefully assessed when interpreting the findings of these epidemiological studies.

72 Confidence Interval An X% (e.g. 90% or 95%) confidence interval has an X% probability of containing the true value that is being estimated (e.g. the risk of a disease). The confidence interval estimate must always be taken into account when interpreting the results of a study.

73 Statistical Power The statistical power of a study is the probability that a real association of a given strength will be detected by the study. Conventionally, it is desirable to conduct a study with a power >80%, if possible. A study with high power will produce narrow confidence interval estimates (i.e. more precise estimates), and vice versa. The precision of an estimate is defined as the reciprocal of the variance (=1/σ 2 ).

74 Epidemiological Evidence Absence of evidence should not be confused with evidence of absence negative results must always be considered in the context of the statistical power of the study, and should not be overzealously interpreted as the absence of an effect. The confidence interval gives the likely range of the strength of effect.

75 Cancer Epidemiological studies of exposed groups have provided clear evidence of an excess risk of cancer following exposure to ionising radiation. Gold Standard is the Japanese survivors of the atomic bombings of Hiroshima and Nagasaki in 1945.

76 Cancer Risk No other environmental carcinogen, with the possible exception of tobacco, has been studied as extensively [as radiation]; yet, there remains a public mystique about radiation that tends to exaggerate the actual hazard. Harvard Report on Cancer Prevention, 1996

77 Measures of Risk Excess Relative Risk (ERR) is the proportional increase in the risk of a disease compared to the background absolute risk in the absence of exposure. For example, an ERR of 1 is a 100% increase of the risk over background. Relative Risk (RR) is the ratio of the overall risk to the background risk (RR = ERR + 1). For example, a RR of 2 is a doubling of the risk. Excess Absolute Risk (EAR) is the additional risk above the background absolute risk in the absence of exposure.

78 Measures of Risk ERR = ( ) / = 2 red / blue RR = / = 3 (red + blue) / blue EAR = = red

79 Exposed Populations Japanese atomic bomb survivors Medically exposed groups Occupationally exposed groups Environmentally exposed groups

80 Hiroshima and Nagasaki 6 th and 9 th August 1945

81 Japanese Atomic Bomb Survivors Cohort study of the Japanese survivors of the atomic bombings of Hiroshima and Nagasaki in 1945 represents the Gold Standard for radiation epidemiology. It is upon the experience of these Japanese survivors that the radiation risk estimates underlying radiological protection are primarily (but not solely) based.

82 Radiation Effects Research Foundation (RERF) RERF is a joint Japanese/US organisation based in Hiroshima and Nagasaki that analyses the atomic bomb survivor data. Updates cancer risk estimates ~2 times a decade.

83 Life Span Study (LSS) Follow-up of ~ survivors having dose estimates; ~ survivors were nontrivially exposed (doses >5 msv), ~⅔ of whom received doses <100 msv. Started in October 1950, still underway. General population of healthy individuals of both sexes and all ages. Mortality and cancer incidence investigated. Wide range of doses received with detailed organ dose estimates (DS02 doses).

84 LSS Leukaemia Mortality Ozasa et al., Radiat Res 2012; 177: covers 318 leukaemia deaths in the Life Span Study (LSS) between 1950 and Based on a linear-quadratic dose-response model (all ages, both sexes): ERR = 3.1 (95% CI: 1.8, 4.3) at 1 Gy ERR = 0.15 (95% CI: -0.01, 0.31) at 0.1 Gy

85 LSS Leukaemia Mortality Preston et al., Radiat Res 2004; 162: covers leukaemia deaths in the LSS between 1950 and deaths, 204 with RBM doses >5 msv, against an fitted background of 203. Approaching half the deaths in the non-trivially exposed are attributed to exposure. Follow up only started in 1950, although it is clear that excess deaths occurred earlier.

86 Leukaemia Deaths ( ) (Preston et al., Radiat Res 2004; 162: ) Sex-averaged EAR in 1970 for age-at-exposure years

87 BEIR VII/NCI Leukaemia Mortality

88 LSS Leukaemia Incidence Hsu et al., Radiat Res 2013; 179: covers leukaemia incidence in the LSS between 1950 and cases (192 with RBM doses >5 msv) against an fitted background of 218. Approaching half the cases in the nontrivially exposed are attributed to exposure. Based on a linear-quadratic dose-response model (all ages, both sexes): ERR = 4.7 (95% CI: 3.3, 6.5) at 1 Gy

89 Leukaemia Incidence ( ) (Hsu et al., Radiat Res 2013; 179: ) Sex-averaged ERR at an attained age of 70 years and an age-at-exposure of 30 years

90 Solid Cancer Mortality ( ) (Ozasa et al. Radiat Res 2012; 177: ) ERR coefficient (Excess Relative Risk per unit dose received the slope of the linear dose-response) for mortality from a solid cancer for a person exposed at age 30 years and at an attained age of 70 years is 0.47 Gy -1 (95% CI: 0.38, 0.56) based upon almost deaths.

91 Excess relative risk (and 95% CI) Gy -1 of mortality from specific solid cancers among the Japanese atomic bomb survivors during (Ozasa et al. Radiat Res 2012; 177: )

92 Solid Cancer Incidence ( ) (Preston et al. Radiat Res 2007; 168: 1-64) The Excess Relative Risk (ERR) of a solid cancer being diagnosed at an attained age of 70 years after receiving a dose of 1 Gy at an age of 30 years (the ERR coefficient for a linear dose-response) is 0.47 Gy -1 (90% CI: 0.40, 0.54) based upon almost cases.

93 Solid Cancer Incidence among Japanese Atomic Bomb Survivors ( ) (Preston et al. Radiat Res 2007; 168: 1-64)

94 Relative Risk of Solid Cancer Incidence among the Japanese Atomic Bomb Survivors during (Pierce and Preston, Radiat Res 2000; 154: ) Relative Risk of Solid Tumour Incidence 1.4 Smoothed fitted line through low dose data 95% confidence band Linear fit through data in 0-2 Sv dose range Gamma-ray Equivalent Dose (Sv)

95 Excess relative risk per Gy (and 90% CI) for the incidence of specific solid cancers among the Japanese atomic bomb survivors during

96 Life Span Study (LSS) Acute, high dose-rate exposure. Malnourished Japanese population; low proportion of men of military age. Some (retrospective) dose estimates uncertain, predominantly external γ doses. Healthy survivor effect follow-up started in October Many survivors still alive at last follow-up. Data prior to October 1950 missing.

97 Radiation Epidemiology Although the Japanese atomic bomb survivors provide fundamental and compelling evidence on the health effects of exposure to radiation it is important to study other groups to produce a broad base of evidence. This is particularly so when the exposure is, for example, protracted rather than brief, or to α-particles rather than γ-rays, i.e. other exposure circumstances of direct relevance to radiological protection.

98 Medically Exposed Groups Ankylosing spondylitis patients Other benign disease patients e.g. ringworm of the scalp, thymus gland enlargement, skin haemangioma, peptic ulcer Cervical cancer patients Other cancer patients Thorotrast (Th-232) patients Radium (Ra-224)-treated patients Diagnostically exposed groups e.g. TB, scoliosis, antenatal X-rays, CT scans

99 Medical Exposures in the USA

100 Medical Radiography A number of studies of exposure from medical radiography indicate that doses of ~10 mgy of X-rays can cause cancer Tuberculosis patients given a series of X-ray examinations show an excess risk of breast cancer proportional to the number of examinations. Scoliosis patients given a series of X-ray examinations show an excess risk of breast cancer proportional to the number of examinations. Case-control studies of obstetric radiography find an excess risk of childhood cancer, particularly childhood leukaemia.

101 Medically Exposed Groups Although medically exposed groups offer a valuable complement to evidence derived from the Japanese atomic bomb survivors care in interpretation is required: Exposure occurs because of known or suspected disease and this may affect the subsequent risk reverse causation and confounding by indication Radiotherapy involves high and localised doses Accurate dose estimates are often lacking

102 CT Scanning

103 CT Scan Cohort Study (Pearce et al., Lancet 2012; 380: ) Cohort study of > patients first examined with CT in Great Britain during when 21 years of age. Cancers diagnosed during identified through UK national cancer registry. Initial analysis of leukaemia and brain tumours, using estimates of RBM and brain doses per CT scan from details contained in medical records.

104 CT Scan Cohort Study Pearce et al., Lancet 2012; 380:

105 British CT Scan Cohort Study (Pearce et al., Lancet 2012; 380: ) Brain Tumours ERR/Gy = 23 (95% CI: 10, 49)

106 Australian CT Scan Cohort Study (Mathews et al., BMJ 2013; 346: f2360) Cohort study of ~11 million Australians. Cancers diagnosed during identified through national cancer records. > patients first examined with CT during when <20 years of age. All cancers (~60 500) included in study. Follow-up commenced 1 year after first CT scan: 3150 cases of cancer identified in CT scan cohort.

107 Australian CT Scan Cohort Study (Mathews et al., BMJ 2013; 346: f2360) Abstract Results The mean duration of follow-up after exposure was 9.5 years. Overall cancer incidence was 24% greater for exposed than for unexposed people. Conclusions The increased incidence of cancer after CT scan exposure in this cohort was mostly due to irradiation. My emphasis

108 Australian CT Scan Study (Mathews et al., BMJ 2013; 346: f2360) No. Exposed Cases IRR (95% CI)

109 Australian CT Scan Study (Mathews et al., BMJ 2013; 346: f2360)

110 Australian CT Scan Study (Mathews et al., BMJ 2013; 346: f2360)

111 Australian CT Scan Study (Mathews et al., BMJ 2013; 346: f2360) The possibility of reverse causation (i.e., that the early symptoms of undetected cancer, or of factors that predispose to cancer, were the indications for the CT scans rather than the CT scans causing the cancers) is certainly important here. The early appearance after first CT scan of solid cancers, and the significant excesses of cancers not thought to be particularly associated with exposure to radiation while there is no excess of radiosensitive breast cancer, reinforces concern over this possibility. It is not clear to us why the authors did not consider these patterns in their results to demand a circumspect interpretation. Walsh, Shore, Auvinen, Jung and Wakeford, BMJ 2013, on-line June

112 Thyroid Cancer (Lubin et al., J Clin Endocrinol Metab 2017, in press) Pooled data from nine cohort studies of childhood external exposure and thyroid cancer, doses <0.2 Gy. For both <0.2 Gy and <0.1 Gy, RRs increased with dose (p<0.01), without significant departure from linearity. Threshold dose >0.04 Gy ruled out. Excess risk arises within 5-10 years and persists for 45+ years.

113 Thyroid Cancer (Lubin et al., J Clin Endocrinol Metab 2017, in press)

114

115 Occupationally Exposed Groups Underground hard rock miners Radium dial painters Radiologists/radiographers Aircrew Nuclear industry workers UK/North America/etc. former USSR (Mayak, Chornobyl emergency and recovery workers)

116 Underground Hard-rock Miners

117 Underground Hard-rock Miners Underground hard-rock miners (e.g. uranium, iron, gold, tin miners) inhale radon (mainly 222 Rn) and its radioactive decay products. In the past, exposures have been high. A clear radon-related excess of lung cancer has been demonstrated in many groups of miners, but little evidence for an excess risk of other cancers associated with exposure.

118 Radon and Lung Cancer (Lubin et al., J Natl Cancer Inst 1995; 87: ) 16 Relative Risk (and 95% CI) of Lung Cancer by Cumulative Exposure to Radon Progeny (Working Level Months, WLM). Combined Data from Eleven Cohorts of Underground Hard Rock Miners. 14 Fitted Linear ERR Model Relative risk of lung cancer Cumulative exposure to radon progeny (WLM)

119 Radon and Lung Cancer (Lubin et al., J Natl Cancer Inst 1995; 87: ) Relative Risk (and 95% CI) of Lung Cancer by Cumulative Exposure to Radon Progeny (Working Level Months, WLM). Combined Data for Cumulative Exposure <400 WLM from Eleven Cohorts of Underground Hard Rock Miners. 4 Fitted Linear ERR Model Relative risk of lung cancer Cumulative exposure to radon progeny (WLM)

120 Radium Dial Luminizers

121 Nuclear Industry Workers

122 Nuclear Industry Workers Several studies of nuclear industry workforces have been carried out. These offer the direct investigation of exposure circumstances of primary interest to radiological protection. Average cumulative doses are low so that large numbers are required for studies to have reasonable statistical power.

123 Nuclear Industry Workforce Studies Workforce studies in countries with nuclear industries established in the 1940s and 1950s are now reaching maturity because early workers with the highest cumulative doses are reaching an age with an appreciable chance of serious illness and death health event data are increasing rapidly. Predicted levels of radiation-related effects among these workforces may be anticipated to be manifest in the near future.

124 NMBU, Ås, Norway, 15 June 2017 Nuclear Industry Workers Sellafield, England Mayak, Russia

125 Doses for Sellafield Workers (Douglas et al., Br J Cancer 1994; 70: ) Average annual external dose per radiation worker

126 Nuclear Worker Studies In general, nuclear industry workers will have accumulated radiation doses over a number of (sometimes many) years. Although some early workers may have received high cumulative doses (some doses ~1 Sv), the pattern is many small doses accumulated at low dose-rates. This pattern of exposure is important because protracted exposure to periods of moderately raised dose-rates is of primary importance to radiological protection.

127 Pattern of Occupational Exposure

128 Mayak Nuclear Complex, Southern Urals, Russia

129 Mayak, Ozyorsk, Chelyabinsk, Russia

130 Mayak Workers The early Mayak workforce experienced high exposures to external radiation and plutonium over a protracted period. Studies of Mayak workers are potentially of appreciable value, but several issues (e.g. dosimetry) require resolution before risk estimates can be relied upon.

131 Mayak Workers (Vostrotin et al., Radiat Prot Dosim 2016; in press) Mean organ dose from external sources of photons: ~0.5 Gy (max. ~8 Gy) Mean lung dose from Pu: 0.19 Gy (max. 9 Gy) Mean liver dose from Pu: 0.18 Gy (max. 11 Gy) Mean bone surface dose from Pu: 0.71 Gy (max. 47 Gy)

132 Mayak Workers Mortality from Solid Cancers, other than Lung, Liver and Bone, in relation to the Cumulative Dose of Recorded External Radiation. (Sokolnikov et al., PLoS ONE (2) e011784) ERR/Gy = 0.16 (95% CI: 0.07, 0.26)

133 Plutonium Exposure at Mayak (Gilbert et al., Radiat Res 2013; 179: ) Lung Cancer Mortality

134 Liver Cancer at Mayak (Sokolnikov et al., Int J Cancer 2008; 123: )

135 Mayak Workers (Sokolnikov et al., Int J Cancer 2008; 123: ) Estimated radiation-induced deaths (percentage of total deaths) Lung Liver Bone Internal: 199 (29%) External: 58 (9%) Internal: 27 (35%) External: 6 (8%) Internal: 14 (47%) External: 3 (10%) NMBU, Ås, Norway, 15 June 2017

136 Leukaemia at Mayak (Kuznetsova et al., PLoS ONE 2016; 11(9): e ) ERR coefficient for leukaemia (excluding CLL) incidence during following external exposure to radiation 3.46 (90% CI: 1.57, 7.65) Gy -1 (2-year dose lag)

137 Leukaemia at Mayak (Kuznetsova et al., PLoS ONE 2016; 11(9): e ) External irradiation leukaemia (excluding CLL) incidence: Years Since Dose Received ERR/Gy (90% CI) (10.87, 47.82) > (-0.11, 2.16) Plutonium exposure leukaemia (excluding CLL) incidence: ERR/Gy = 3.63 (90% CI: <0, 15.85)

138 International Collaboration During the 1980s, discussions were held to consider the possibility of an international study of nuclear industry workers, which would increase statistical power. These discussions resulted in the International Collaborative Study of Cancer Risk among Radiation Workers in the Nuclear Industry, coordinated by the International Agency for Research on Cancer (IARC) based in Lyon, France.

139 3-Country Study (Cardis et al., Radiat Res 1995; 142: ) First IARC study considered nuclear workers from three countries UK UKAEA, AWE, BNFL Sellafield USA DOE sites at Hanford, Oak Ridge, Rocky Flats Canada Atomic Energy of Canada Ltd (AECL)

140 3-Country Study (Cardis et al., Radiat Res 1995; 142: ) 95,673 radiation workers Average individual dose 40.2 msv, collective dose 3,843.2 person.sv Average follow-up 22.2 years 3,976 cancer deaths

141 3-Country Study (Cardis et al., Radiat Res 1995; 142: ) Comparison with Japanese atomic bomb survivors (ERR, Excess Relative Risk, is the proportional increase in risk) Leukaemia (except CLL) IARC ERR = 2.18 (90% CI: 0.13, 5.7) per Sv A-bomb ERR = 3.67 (90% CI: 2.0, 6.5) per Sv All cancers except leukaemia IARC ERR = (90% CI: -0.39, 0.30) per Sv A-bomb ERR = 0.18 (90% CI: 0.05, 0.34) per Sv

142 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) Second IARC study considered nuclear workers from 15 countries: Australia Belgium Canada Finland France Hungary Japan South Korea Lithuania Slovak Republic Spain Sweden Switzerland UK USA

143 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) After exclusions, ~ workers included. ~ deaths, of which 6519 solid tumours and 196 leukaemias (excluding CLL). ~5 million person-years of follow-up, average follow-up years. Collective external radiation dose of 7892 person-sv, a mean individual cumulative dose of 19.4 msv.

144 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) ~ workers initially excluded from the study for various reasons. In particular, ~ exclusions due to potential neutron exposure and ~ exclusions due to potential exposure to radionuclides (although there is overlap between these groups). Unfortunately, these exclusions also removed high cumulative gamma-ray dose workers and led to loss of statistical power.

145 15-Country Study (Cardis et al., Radiat Res 2007; 167: ) Only 13 cancer deaths occurred in the 400+ msv external dose group (and only 6 cancer deaths in the 500+ msv dose group). Workers excluded because of potential neutron or internal exposures had a higher average cumulative gamma-ray dose (46.6 msv) than workers included in the study (19.4 msv), which is indicative of a substantial loss of statistical power.

146 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) Further exclusions were made for some analyses. For cancers other than leukaemia, 413 worker deaths from Japan, 196 from Ontario Hydro (Canada) and 886 from Idaho Falls (USA) were excluded because socioeconomic [status] information was unavailable or incomplete (although the exclusion criteria were not given). Further loss of power (and possible bias?).

147 SMR* by Cohort Cohort Mortality from All Causes Mortality from All Cancers Obs SMR* 95% CI Obs SMR* 95% CI Australia Belgium Canada 1, Finland France (CEA-COGEMA) France (EDF) Hungary Japan 1, Korea (South) Lithuania Slovak Republic Spain Sweden Switzerland UK 7, , U.S. (Hanford) 5, , U.S. (INL) 3, U.S. (NPP) U.S. (ORNL) 1, *Standardised Mortality Ratio: ratio (%) of observed number of deaths (Obs) to number expected from national statistics.

148 Healthy Worker Effect Consistently low SMRs show a strong healthy worker effect, although some degree of under-ascertainment could also be present. Healthy worker effect could vary with length of employment, requiring adjustment by duration of employment. Place most evidential weight upon trends with cumulative external radiation dose.

149 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) ERR (Excess Relative Risk) coefficients Leukaemia (minus CLL) 1.93 (95% CI: <0, 8.47) Sv -1 (2-year lag) Cancers other than leukaemia 0.97 (95% CI: 0.14, 1.97) Sv -1 (10-year lag)

150 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) Cancers other than leukaemia 0.97 (95% CI: 0.14, 1.97) Sv -1 Lung cancer 1.86 (95% CI: 0.26, 4.01) Sv -1 Cancers other than leukaemia and cancers of the lung and pleura 0.59 (95% CI: -0.29, 1.70) Sv -1

151 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) Smoking-related cancers other than lung cancer 0.21 (95% CI: <0, 2.01) Sv -1 Cancers unrelated to smoking 0.62 (95% CI: -0.51, 2.20) Sv -1

152 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) All non-malignant respiratory diseases 1.16 (95% CI: -0.53, 3.84) Sv -1 Chronic obstructive bronchitis and emphysema 2.12 (95% CI: -0.57, 7.46) Sv -1

153 15-Country study (Cardis et al., BMJ 2005; 331: 77-80) Taken together, these findings indicate that a confounding effect by smoking may be partly, but not entirely, responsible for the estimated increased risk for mortality from all cancers other than leukaemia. Cardis et al. (2005)

154 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80) The unresolved question is to what degree confounding by tobacco smoking is influencing the ERR coefficient for cancers other than leukaemia, i.e. to what extent must this ERR/Sv be adjusted to take account of the effect of smoking?

155 15-Country Study (Cardis et al., BMJ 2005; 331: 77-80)

156 15-Country Study (Cardis et al., BMJ 2005; 331: 77-82) 204 deaths from cancers other than leukaemia are from Canada (4% of total). Exclusion of Canadian data leads to a 40% reduction in ERR coefficient from to 0.97 (95% CI: 0.14, 1.97) Sv (95% CI: -0.22, 1.55) Sv -1

157 15-Country Study (Cardis et al., Radiat Res 2007; 167: ) 65 deaths from lung cancer are from Canada (4.5% of total). Exclusion of Canadian data leads to a 50% reduction in ERR coefficient from to 1.86 (90% CI: 0.49, 3.63) Sv (90% CI: -0.39, 2.62) Sv -1

158 15-Country Study (Cardis et al., Radiat Res 2007; 167: ) Canadian ERR Coefficients All cancers excluding leukaemia Lung cancer 6.65 (90% CI: 2.56, 13.0) Sv (90% CI: 3.63, 27.8) Sv -1

159 15-Country Study (Cardis et al., Radiat Res 2007; 167: ) All cancers excluding leukaemia (ERR/Sv) Canada 6.65 (90% CI: 2.56, 13.0) Sv -1 Other 14 countries combined 0.58 (90% CI: -0.10, 1.39) Sv -1

160 Canadian Worker Studies (Ashmore et al., J Radiol Prot 2010; 30: 121-9) country minus Canada

161 Canadian Nuclear Workers (Zablotska et al., Br J Cancer 2014; 110: ) Significantly increased risks for early AECL workers are most likely due to incomplete transfer of AECL dose records to the National Dose Registry. Analyses of the remainder of the Canadian nuclear workers (93.2%) provided no evidence of increased risk, but the risk estimate was compatible with estimates that form the basis of radiation protection standards.

162 Canadian Nuclear Workers (Zablotska et al., Br J Cancer 2014; 110: ) Study findings suggest that the revised Canadian cohort, with the exclusion of early AECL workers, would likely have an important effect on the 15-country pooled risk estimate of radiation-related risks of all cancer excluding leukaemia by substantially reducing the size of the point estimate and its significance.

163 Canadian Nuclear Workers (Zablotska et al., Br J Cancer 2014; 110: ) Leukaemia (ERR/Sv) 14.4 (95% CI: <-1.49, 146) All cancers excluding leukaemia (ERR/Sv) (95% CI: <-1.47, 1.98)

164 15-Country Study Until the Canadian (especially the AECL) data have been examined carefully, it would be prudent to consider the results with the Canadian data excluded from the analyses All cancers excluding leukaemia Lung cancer 0.58 (90% CI: -0.10, 1.39) Sv (90% CI: -0.39, 2.62) Sv -1

165 IARC 15-country Study (Cardis et al., BMJ 1995; 331: 77-82) Unfortunately, there are difficulties in the interpretation of the findings of this study Influenced by lung cancer and confounding by tobacco smoking Influenced by the Canadian data, which look suspicious and require investigation Excluded workforces could have introduced bias Many high cumulative external dose workers excluded, which reduces statistical power

166

167 INWORKS The International Nuclear Workers Study (INWORKS) includes workers from three countries (France, UK, USA). ~ radiation workers included, with ~8.2 million person-years of follow-up. Average cumulative dose, ~21 mgy. Collective dose, ~5400 person.gy. ~ workers with doses >100 mgy. ~ cancer deaths.

168 INWORKS Error bars show 90% confidence intervals. (Leuraud et al., Lancet Oncol 2015; 2: e276-81)

169 INWORKS Error bars show 90% confidence intervals. (Richardson et al., BMJ 2015; 351: h5359)

170 INWORKS (Richardson et al., BMJ 2015; 351: h5359) (Leuraud et al., Lancet Oncol 2015; 2: e276-81) Leukaemia excluding CLL INWORKS: ERR/Gy = 2.96 (90% CI: 1.17, 5.21) LSS: ERR at 1 Sv = 2.63 (90% CI: 1.50, 4.27) All solid cancers INWORKS: ERR/Gy = 0.47 (90% CI: 0.18, 0.79) LSS: ERR/Sv = 0.32 (95% CI: 0.01, 0.50)

171 INWORKS Error bars show 90% confidence intervals. (Leuraud et al., Lancet Oncol 2015; 2: e276-81)

172 INWORKS Error bars show 90% confidence intervals. (Richardson et al., BMJ 2015; 351: h5359)

173 INWORKS Questions At present, INWORKS only considers photon doses, not neutron or internal doses. Early missed photon doses? What is the impact on ERR/Gy estimates?

174 Sellafield, Cumbria, UK

175 NRRW-3 (follow-up to end of 2001) The number of workers with lifetime doses 100 msv, accumulated over many years, is substantial: Only ~¼ of workers with cumulative doses 100 msv had died by the end of (~ individuals in the Life Span Study received doses 100 msv)

176 Environmentally Exposed Groups High natural background external gamma radon Weapons testing fallout Utah, Marshall Islands, Nordic countries, former USSR (e.g. Semipalatinsk) Contamination Chornobyl, former USSR (e.g. Techa River), Hanford, Taiwanese steel, Fukushima

177 Naturally Occurring Sources of Radiation Cosmic radiation from the Sun and beyond direct external exposure ( cosmic rays ) intakes of radionuclides ( 3 H, 14 C) produced in the upper atmosphere Terrestrial radiation from long-lived radionuclides and their decay products direct external exposure intakes of radionuclides ( 222 Rn/ 220 Rn in air; 238 U, 232 Th, 226 Ra, 210 Po, 210 Pb, 40 K in food and drink)

178 Average Annual Radiation Dose The average annual effective dose received by an individual living in the UK is 2.6 msv, 2.2 msv of which is from natural sources. (These are early 2000s values medical exposure will have increased). Around the world and within countries there are large variations in radiation doses from natural background radiation, largely from radon variations, but also from γ-radiation from naturally-occurring sources in the environment.

179 Residential Radon Case-control (i.e. individual- rather than group-based) studies of residential radon exposure and lung cancer take account of both radon exposure and smoking histories. Appropriately pooled data from case-control studies in Europe, North America and China find associations between domestic exposure to radon and lung cancer.

180 Residential Radon (Krewski et al., Epidemiol 2005; 16: ) Pooled data from 7 North American casecontrol studies gives an odds ratio for lung cancer risk with radon concentration of 1.11 (95% CI: 1.00, 1.28) at 100 Bq/m 3 When data were restricted to radon measurements for those residing in just one or two houses the odds ratio becomes 1.15 (95% CI: 1.01, 1.37) at 100 Bq/m 3 Miner studies predict 1.12 (95% CI: 1.02, 1.25) at 100 Bq/m 3

181 Residential Radon (Krewski et al., Epidemiol 2005; 16: ) AAAAAAAA Radon concentration (Bq/m 3 )

182 Residential Radon (Darby et al., BMJ 2005; 330: 223-8) Pooled data from 13 European casecontrol studies gives an odds ratio for lung cancer risk with radon concentration of 1.08 (95% CI: 1.03, 1.16) per 100 Bq/m 3 When corrected for random uncertainties in radon measurements this becomes 1.16 (95% CI: 1.05, 1.31) per 100 Bq/m 3

183 Residential Radon (Darby et al., BMJ 2005; 330: 223-8) Relative Risk of Lung Cancer with respect to the Estimated "Usual" Residential Radon Concentration. Combined Data from 13 European Case-Control Studies. Error Bars show 95% Floated Confidence Intervals Fitted Linear ERR Model Relative Risk (95% floated CI) Estimated "Usual" Radon Concentration (Bq/m 3 )

184 Residential Radon (Darby et al., BMJ 2005; 330: 223-8)

185 High Natural Background Radiation Areas (HNBR areas) Guarapari, Brazil; Kerala, India; Ramsar, Iran; Yangjiang, China, are all recognised HNBR areas that have been investigated to varying extents. Kerala and Yangjiang have been paid particular attention.

186 Yangjiang, Guangdong, China

187 Radiation Exposure in Yangjiang

188 Yangjiang HNBR Area, China Tao et al. (Health Phys 2012; 102: ) examined mortality in residents aged years during Cumulative gamma-ray doses by village deaths (956 cancers) studied. No significant correlations, except a negative correlation between liver cancer mortality and cumulative dose.

189 Kerala, Southern India

190 Karunagappally Taluk (Hosoda et al., PLoS ONE (4) e ) External exposure to γ-rays from thoriumbearing monazite sands.

191 Kerala HNBR Area, India (Nair et al., Health Phys 2009; 96: 55-66) Nair et al. examined cancer incidence (from a regional cancer registry) among 69,958 residents aged years during > gamma dose-rates measured cases of cancer (30 leukaemia). No significant correlations. Cancer excluding leukaemia: ERR/Gy = (95% CI: -0.58, 0.46)

192 Cancer Risk in Kerala (Nair et al., Health Phys 2009; 96: 55-66) NMBU, Ås, Norway, 15 June 2017

193 Cancer Risk in Kerala

194 Kerala, India

195 Bidi Smoking in Kerala

196 Natural Background Radiation Perhaps a more focused investigation of the risk model-predicted effects of natural background radiation would be more informative? Childhood leukaemia has a predicted high ERR/Sv, but above a low background risk (1 in 1800 live births affected in economically developed countries).

197 BEIR VII/NCI Leukaemia Mortality

198 Natural Background Radiation (Wakeford et al., Leukemia 2009; 23: Little et al. J Radiol Prot 2009; 29: Kendall et al., Leuk Res 2011; 35: ) Recent risk models for radiation-induced leukaemia suggest that ~15% of cases of childhood (<15 years of age) leukaemia in Great Britain may be caused by natural background radiation. red bone marrow dose ~1.4 msv per annum Past epidemiological studies have been unable to reliably demonstrate this source of risk probably have insufficient statistical power

199 Natural Background γ-radiation (Kendall et al., Leukemia 2013; 27: 3-9) Case-control study of 27,500 childhood cancer cases (of which, 9000 leukaemia cases) and 37,000 controls in Great Britain, Cumulative dose of γ- rays from birth to diagnosis, by county district of birth. Childhood Leukaemia Childhood Cancers Except Leukaemia ERR/Gy = 120 (95% CI: 30, 220)

200 Nuclear Weapons Testing

201 Cs-137 and Pu in Fallout (Warneke et al., Earth Planet Sci Lett 2002; 203: )

202 Weapons Testing Fallout

203 Childhood Leukaemia Incidence (Wakeford et al. Radiat Environ Biophys 2010; 49: 21-27) 100 Incidence Rate of All Leukaemias (Except Where Indicated Otherwise) among Children Aged 0-14 Years, Incidence Data from Eleven Cancer Registries. Error Bars Show 95% Confidence Intervals for Rates. Registartion Rate of Leukaemia in the 0-14 Year Age Group, cases per million person-years Connecticut, Saskatchewan, New Zealand, Great Britain, Denmark, Sweden, Norway, Finland, Hawaii, Baltimore (AL), Western Australia (ALL), Calendar Year of Diagnosis

204 Chornobyl 26 April 1986

205 Chornobyl Contamination

206 Chornobyl Thyroid Cancer (Yamashita, Health Phys 2014; 106: )

207 Chornobyl Thyroid Cancer Substantial numbers of children in some areas of the former USSR received high thyroid doses (up to 1 Sv or more) as a result of the release of radioiodine during the Chornobyl reactor accident. >4000 excess thyroid cancers have occurred with the possibility of many more to come.

208 Childhood Thyroid Cancer (Ron et al., Radiat Res 1995; 141: ) Atomic bomb survivors and therapeutically irradiated patients demonstrate that the thyroid glands of children are very sensitive to radiation-induced cancer. Ron et al. (1995) conducted a pooled analysis of seven externally irradiated groups. For those exposed before the age of 15 years the ERR coefficient was 7.7 Sv -1 (95% CI: 2.1, 28.7).

209 Chornobyl Risk estimates for thyroid cancer are broadly compatible with predictions from external irradiation studies. Little firm evidence for excesses of other childhood cancers (including childhood leukaemia) or adult cancers (including thyroid cancer) being associated with Chornobyl contamination, although studies continue.

210 Age Standardised Mortality from All Causes by District of Russia (Men et al. BMJ 2003; 327: 964)

211 All Cause Mortality in Russia in the Year Age Group (Zaridze et al. Lancet 2009; 373: )

212 Chornobyl Liquidators

213 Fukushima Dai-ichi

214 Nuclear Reactor Accidents A comparison of the activities (PBq) of radionuclides released to atmosphere as a consequence of the Fukushima and Chornobyl accidents.

215 Nuclear Weapons Testing Comparison of Activity Releases (PBq) from Atmospheric Nuclear Weapons Testing and the Chornobyl Accident Radionuclide Nuclear Weapons Testing (PBq) Chornobyl Accident (PBq) I Cs Sr Pu (α-activity)

216 Fukushima Contamination

217 Thyroid Cancer in Fukushima (Tsuda et al. Epidemiol 2015 in press) ~ thyroid ultrasound examinations of Fukushima Prefecture residents 18 years Number of thyroid cancers detected by screening programme was 110, ~30 times more than might be expected from Japanese incidence rates However, no significant variation of prevalence across the prefecture, despite heterogeneous levels of contamination

218 Thyroid Cancer, South Korea (Ahn et al. New Eng J Med )

219 Thyroid Cancer, South Korea (Ahn et al. New Eng J Med )

220 Thyroid Cancer Chornobyl vs Fukushima

221 Risks to General Population (WHO Expert Group, 2013) For an infant in the most affected location, the ERR of leukaemia over a lifetime is 7%; of breast cancer, 4%; of thyroid cancer, 70%; and of all solid cancers, 4%. The EAR of leukaemia over a lifetime (the LAR) is <0.05%; of breast cancer, <0.5%; of thyroid cancer, ~0.5%; and of all solid cancers, ~1%. The risks are less than this for other ages at exposure and other locations.

222 Hanford, Washington State

223 I-131 and Hanford (Davis et al., JAMA 2004; 292: ) 27 PBq I-131 released to atmosphere from Hanford during Davis et al. (2004) conducted a historical cohort study of nearly 3500 people born near Hanford during , with individual dose estimates; mean thyroid dose 174 mgy. Non-significant ERR coefficient of 0.7 Gy -1.

224 Taiwanese Contamination (Hwang et al., Radiat Res 2008; 170: ) Steel contaminated with 60 Co was used as a construction material in Taiwan during Affected buildings identified in Hwang et al. (2008) studied just over 6000 people identified as having been exposed (average cumulative dose ~50 msv). Raised rates of leukaemia (excluding chronic lymphocytic leukaemia) and breast cancer.

225 1957 Kyshtym Accident

226 Techa River Contamination

227 Techa River

228 Leukaemia at Techa River (Krestinina et al., Br J Cancer 2013; 109: ) ERR/Gy = 2.2 (95% CI: 0.8, 5.4)

229 Solid Cancer at Techa River (Davis et al., Radiat Res 2015; 184: 56-65)

230 Factors Affecting Risk Dose Dose-rate Sex Age at exposure Time since exposure and/or Attained age

231 Types of Cancer Tissues vary in susceptibility to radiationinduced cancer. Some types of cancer (e.g. chronic lymphatic leukaemia, Hodgkin s lymphoma, malignant melanoma of the skin) do not appear to be induced by ionising radiation, or have low sensitivity.

232 Dose-response Model Conventionally, for radiological protection For leukaemia, an explicit no-threshold linearquadratic model is adopted. For solid tumours, for sparsely ionising radiation (but not for densely ionising radiation), an implicit no-threshold linear-quadratic model is adopted with the slope at low doses being half that at high doses a dose and dose-rate effectiveness factor (DDREF) of 2.

233 Risk of Low-level Exposure Is there a risk of adverse health effects (e.g. cancer) following low-level exposure to ionising radiation? If so, are the risk estimates currently in use for radiological protection purposes appropriate for this use?

234 Ionising Radiation and Cancer It is beyond rational dispute that exposure to moderate and high levels of ionising radiation increases the risk of most forms of cancer in the exposed individual. The epidemiological and experimental evidence for this is overwhelming. Ionising radiation is an IARC Group 1 carcinogen ( carcinogenic to humans )

235 Low-level Exposure Risks The dose-response relationship at low levels of exposure is based upon : the nature of the relationship at moderate-to-high doses mainly external gamma/x-ray irradiation experienced briefly combined with an incomplete knowledge of biological mechanisms at low doses/dose-rates

236 Dose-response Curves Radiation-induced Risk Some possible dose-response curves describing the excess risk of stochastic health effects at low doses of radiation Linear Sub-linear Threshold Hormesis Supra-linear Bimodal Schematic Data Points (0, 0) Radiation Dose

237 The Discussion of Risks at Low Doses is Characterised by Much Heat and Little Light Those adopting extreme positions are likely to be fixed in their views, which they express forcefully.