INTERNATIONAL SYMPOSIUM ON BIOLOGICAL EFFECTS OF LOW DOSE RADIATION - Molecular mechanisms for radiation-induced cellular response and cancer development, October 9-11, 2002, Rokkasho, Aomori 039-3212, Japan - Organized by Institute for Environmental Sciences (IES). Proceedings published in March 2003. Low-dose radiation carcinogenesis in mammals : A review of A Database of Cancer Induction by Low-Dose Radiation in Mammals: Initial Observations and comparison with some human data Philippe Duport† International Centre for Low-Dose Radiation Research, Institute of the Environment, University of Ottawa, Canada (Statistical analyses: D. Krewski (University of Ottawa), K. Crump (The K.S. Crump Group) and J. Zielinski (Health Canada) † Author f or correspondence (e-mail: firstname.lastname@example.org) ABSTRACT The International Centre for Low-Dose Radiation Research (ICLDRR) has assembled, in a single database, all published results from experimental radiation carcinogenesis in mammals. The database contains experimental conditions and outcomes for about 87,000 exposed animals (˜ 60,000 cancers) and 40,000 controls (˜ 19,000 cancers). Experiments were conducted with all types of ionizing radiation, at doses starting at 10 mGy for gamma radiation, 40 mGy for X- rays, 2 mGy for beta radiation, 2 mGy for alpha radiation, and 5 mGy for neutrons. The data form 748 datasets, each providing a dose-response curve for a particular species, strain, sex, age at exposure exposed to a range of doses under specific exposure conditions. No cancers were observed in the control groups of about 30% of the datasets. When cancers were observed in control animals, no effect or an apparent reduction in cancer risk were observed in 40% of the neutron datasets, 50% of the X-rays datasets, 53% of the gamma datasets, and 61% of the alpha datasets. Apparent reductions in cancer rate, significant at up to 10 standard deviations were observed in mice exposed to 100 and 250 mGy of gamma radiation. It is confirmed that, in some experiments, exposed animals live considerably longer (up to 40%) than their controls. The frequency of lack of effect, protective effects or increased longevity in exposed animals challenges the general validity of the Linear No Threshold hypothesis (LNT). For comparison, clear beneficial effects observed in the US shipyard nuclear workers and anomalies in uranium mine dosimetry leading to overestimated radon risk are also examined. 1. MATERIALS AND METHODS 1.1. Sources Publications that provide information on low-dose radiation carcinogenesis in mammals were identified in the International Radiobiology Archives of Long-Term Animal Studies [1, 2] IRA 1996; 2 Gerber et al. 1999]. However, since the IRA refers to data published until 1996, additional literature searches were conducted to complement the information contained in the IRA and to cover the time period ending in September 2000. 1.2. Data characteristics and data selection 2 The main objective of this work being to study the shape of the dose-response curve for the induction of radiogenic cancer at the lowest doses used in animal, the available literature was reviewed to identify those data that met appropriate inclusion criteria. A primary criterion for taking data into consideration is that organ or body dose is expressed in terms of absorbed dose. That criterion is always met in the case of external radiation (gamma, X-rays, neutrons) whereas the “dose” of internally deposited radionuclides is, in some publications, expressed in terms intake quantity, for example microcuries or becquerels per unit weight, with no mention of the total corresponding absorbed dose accumulated in target organs. Data from such publications are not taken into consideration. A particular study is considered to address low-dose effects when a) the animals were exposed to doses less than 1 Gy and their lifespan was shortened by less than 10%, compared to that of their controls; or b) the life of the exposed animals was not shortened by more than 10%, regardless of the total accumulated dose; or c) There was no life shortening and no cancers in exposed animals, regardless of the incidence of cancer in their controls and the total accumulated dose. Experiments that met the above criteria were considered to be low-dose experiments, although criteria b) and c) lead to including data obtained at doses much greater than 1 Gy. A detailed description of the database, along with some 109 references, has been published elsewhere . The database contains information on about 87,000 exposed animals, 40,000 controls (Table 1). Dose ranges and the distribution of datasets between radiation types are given in Table 2. As seen in Table 3, animals were exposed to external radiation in more than 70% of the experiments and to internally deposited radioactive elements in the remainder 30%. Table 1. Summary of collected data Type of Radiation Number of Animals Total Alpha Neutrons X-rays Gamma Beta Controls 7,200 7,935 8,083 11,877 5,196 40,291 Cancers in Controls 1,110 5,830 2,411 8,852 786 18,989 Exposed 10,848 23,525 13,095 29,294 10,886 87,648 Cancers in Exposed 3,746 19,617 7,565 26,133 2,007 60,068 Table 2. Number of datasets and dose range by type of radiation. Type of Number of % Min. Dose (Gy) Max. Dose (Gy) Radiation Datasets Gamma 139 18.6 0.002 3.10 Low LET X-ray 145 19.4 0.005 3.32 Beta 79 10.6 0.04 4.00 Alpha 143 19.1 0.1 3.29 High LET Neutron 242 32.4 0.002 8.95 3 Table 3. Mode of radiation deposition Number of Mode of Deposition Percentage (%) Datasets Internal Inhalation 81 10.8 Injection 120 16.0 Ingestion 12 1.6 Instillation 3 0.4 External 532 71.1 Total 748 100.0 An important parameter in the determination of U -shaped dose responses is the cancer rate in control animals (thereafter called “Ic” or “background”). In the collected data, no cancers were observed in controls animals for 24.9% of the datasets, which totally precludes the observation of a possible hormetic effect (Table 4). There are also 11.4% of the datasets in which the cancer rate is less than 1%, which brings the total percentage of datasets in which U-shaped dose responses are impossible or difficult to observe to 36.%. Table 4. Distribution of datasets by category of cancer proportio n in control group. Cancer rate Alpha Beta Gamma Neutron X-ray Total in controls N % N % N % N % N % N % Ic = 0 73 51.0 30 38.0 0 0.0 32 13.2 50 34.5 185 24.9 0 > Ic = 1% 17 11.9 5 6.3 6 4.3 33 13.6 24 16.6 85 11.4 1%>Ic=10% 37 25.9 31 39.2 61 43.9 87 36.0 32 22.1 248 33.2 10%>Ic =25% 7 4.9 6 7.6 39 28.1 42 17.4 22 15.2 116 15.5 Ic>25% 9 6.3 7 8.9 32 23.0 48 19.8 17 11.7 113 15.1 total 143 100.0 79 100.0 139 100.0 242 100.0 145 100.0 748 100.0 2. RESULTS 2.1. Effect of radiation exposure on the lifespan of exposed animals In some, but not all publications, the survival time of animals is reported either as the mean or the median survival time of control and expose animals, or as the survival time in control and exposed animals that died from a specific disease. It is therefore difficult to compare survival data between publications. A general comparison of lifespan between control and exposed animals is given in Table 5. The lifespan of exposed animals exceeds that of controls in about 30% to 46% of the dose levels, depending on the type of radiation. The observed life prolongation can be quite substantial, for example; For example, the mean survival time in mice with alveolar carcinoma following a neutron dose of 250 mSv  exceeds that of controls by 41% (Fig. 1). 4 Table 5. Comparison of mean survival time in control and exposed animals Dose range Datasets with prolonged life Maximum life prolongation (Gy) (%) (%) Gamma 0 – 0.50 37 % 50% Alpha 0 – 1.00 29% 76% Neutrons 0 – 0.50 45% 54% 2.2. Dose and cancer incidence An increase in cancer incidence in exposed animals can be seen whether or not cancers are observed in control animals, but a reduced cancer rate can be seen only if cancers are present in control animals. Therefore, the datasets are separated in five groups, Ic=0, 0<Ic<1%, 1<Ic<10%, 10<Ic<25%, Ic>25% in order to put radiation carcinogenesis and U-shapedness in perspective with cancer rates in control animals (Table 4). MST (days) 1500 Relative risk 1.05 1000 1 -9 sd 0.95 500 Dose (Gy) 0.9 0 -11 sd Dose (Gy) 0.85 0 0.2 0.4 0.6 0.8 1 0.00 0.20 0.40 0.60 Fig. 1 Mean survival time vs dose in 21- Fig. 2 Relative risk for all cancers day old mice with alveolar carcinoma after in mice exposed to gamma radiation neutron exposure (squares: exposed, circles: (reticular tissue and solid cancers)  controls)  2.3. Difficulties in evaluating evidence for U-shapedness using animal data Unambiguous U-shaped dose responses are found in some datasets (for example  and Figure 2). However, in a meta-analysis it is important to address, not simply whether or not individual dose responses can be found that demonstrate statistically significant evidence for U- shapedness, but whether these U-shaped dose-response curves occur with a frequency that is too large to be attributable to chance. In examining the existence of U-shapes dose response in irradiated animals, it is also important to keep in mind that the aim of most, if not all the experiments was to detect and measure a radiation risk, that is, to measure an increase in cancer rate with increasing dose. The consequences of that specific aim on experimental design are that: a) Most experiments were not designed to detect U -shapedness and have no substantial numbers of animals exposed in the dose range where U-shapedness may be occurring; and b) The size of the effect may be small. The maximum reduction in response represents generally only a 30-60% change from the response in controls. Moreover, as indicated above, a low background rate also makes it difficult to detect a U - shaped dose-response. U-shaped dose responses can be observed and analyzed only if a No Observed Effect Level (NOEL) exists in a dataset. There are 472 such datasets in the database. As can be intuitively expected, the probability of observing U-shaped doses responses should increase both with increasing cancer rate in control animals and with the number of dose levels between zero and the 5 NOEL. This seems to be confirmed by the data given in Table 7. The percentage of datasets with apparent U-shaped dose responses reaches more than 40% when at least three dose levels are below the NOEL. Table 6. Percent of datasets with statistical evidence for U-shapedness by number of dose levels below NOEL and background response (P-value ≤ 0.05) Number of Dose Levels Below NOEL Background Response 0 1 2 3 % (N) % (N) % (N) % (N) >0 9 (472) 11 (354) 14 (209) 21 (96) = 5% 16 (255) 19 (186) 26 (101) 38 (48) = 10% 21 (187) 21 (136) 32 (78) 42 (41) = 20% 24 (115) 30 (84) 35 (54) 41 (32) The apparent existence of U -shaped dose responses seems to be independent of the species exposed (Table 7) and of the cancer classification (Table 8). Table 7. Percent of datasets with statistical evidence for hormesis by animal species and sex (P-value ≤ 0.05) Background > 0% Background = 5% Species & Sex % (N) % (N) Mouse 9 (354) 16 (188) Males 10 (136) 14 (71) Females 9 (213) 17 (115) Rat 5 (56) 9 (32) Males 8 (13) 100 (1) Females 5 (37) 7 (31) Dog 13 (62) 23 (35) Table 8. Percent of datasets with statistical evidence for U- shapedness by cancer types (P-value ≤ 0.05) Background > 0% Background = 5% Cancer type % (N) % (N) Carcinoma 8 (112) 13 (62) Leukemia 6 (53) 11 (27) Lymphoma 11 (55) 16 (37) Sarcoma 10 (72) 44 (16) Carc. & sarc. 25 (24) 31 (13) Others 8 (156) 13 (100) 2.4. Human data A large number of populations of radiation workers have been subject to epidemiological studies, but their results and conclusions are difficult to interpret because socioeconomic, lifestyle, access to medical support and prevention are different in the exposed and reference populations. These are sources of bias when comparing mortality or morbidity rates between disparate populations. However, there are at least two populations of radiation workers for which 6 the exposed and reference populations are directly comparable; they are the British radiologists and the US shipyard workers. In the British radiologist study mortality rates from various causes were compared between radiologists and non-radiologist medical personnel. In the US shipyard workers, mortality rates from various causes were compared between nuclear and non-nuclear shipyard workers. In these two populations, reduced mortality rates in exposed workers cannot be attributed to the so-called “healthy worker effect” because the exposed and reference populations are as comparable as can be, except for there occupational exposure to radiation. Furthermore, several populations of uranium miners have also been subject to epidemiological studies. These populations are particularly useful in the study of the risk of cancer posed by inhaled radon decay products because they seem to display a clear increase in risk with increasing exposure (dose). In fact, it seems that an important contribution of radiation sources other than radon decay products to the dose received by the lung, the only organ at risk in uranium miners. This results in a systematic overestimation of the risk per unit exposure (dose) of radon decay products and increase the likelihood of no effect at relatively low indoor radon concentrations. Cancer rates in exposed and reference populations of British radiologists and US shipyard workers, and the overestimation of the risk from inhaled radon decay products are examined below. 2.5. British radiologists The mortality rates in British radiologists have been compared to those in non-radiologist medical personnel over more than one century . The main findings of this study are summarized in Table 9. A decreasing trend in the SMR for all cancers, all non-cancer causes and all causes is observed from 1897, the beginning of medical radiology to 1979. This trend probably corresponds to the progressive implementation of sound radiation protection practices in the profession. Table 9. Evolution of the SMR for all non-cancer, all cancer and all causes in British radiologists (Controls : all male medical practitioners). Year of registration Cause of death 1897 - 1920 1921 - 1935 1936 - 1954 1955 - 1979 O E SMR O E SMR O E SMR O E SMR All non-cancer 230 266 0.86 219 255 0.86 278 291 0.96 77 121 0.64 All cancers 60 34 1.76 51 41 1.24 85 76 1.12 32 45 0.71 All causes 290 300 0.97 271 296 0.92 368 367 1.00 113 166 0.68 2.6. Us shipyard workers The US Department of Energy published in 1994 a large study on mortality rates from a variety of causes among the shipyard workers, some of who were assigned, in comparable number, to nuclear vessels and others to conventional vessels . Like in many reports of that nature, the results are given in tables and actual differences in mortality rates are difficult to visualize. This has been done for mortality rates from all causes, all neoplasms, all cardiovascular diseases and infectious diseases in the shipyard workers (Figure 3). The risk of dying from any of the above cases (at the exception of infectious diseases, for which the number of cases is small) is significantly lower in nuclear workers. 7 under5mSv 1.6 over5mSv 19 cases AllNW 1.4 2 cases 21 cases 1.2 OR=1 1.0 O d r to d s ai 0.8 0.6 0.4 0.2 0.0 Allcauses Neopl Infect CardVasc Cause of death Fig. 3 Odds ratios for all causes of death, all neoplasms, all infectious diseases and all cardiovascular diseases in the US nuclear shipyard workers (Non-nuclear workers are the reference population, 95% confidence intervals). 2.7. Cancer risk for the lung and other organs in uranium miners Since the first epidemiological study of lung cancer in uranium miners, it has been assumed that the dose from gamma radiation and inhaled radioactive dust was negligible compared to the dose from radon progeny. This may have been due to the fact that, as recommended by the ICRP and required by regulatory authorities occupational doses be recorded in terms of effective doses, which concern the whole body. In the case of uranium miners, however, the lung is the only organ at risk . Therefore, it is the dose to the lung, not the effective dose that counts in risk estimation. It must also be noted that the radiation weighting factor, wR, for alpha radiation is most likely not 20 at low doses and dose rates as demonstrated in human data [for example 9, 10] and animal studies [for example 11, 12]. From that viewpoint, lung doses from gamma and radioactive dust are not negligible and may exceed the dose from radon decay products . A review of doses from every source has been conducted in miners equipped with continuous personal monitors for radon decay products. Individual annual doses from each source are summarized in Table 10. The relative contribution of radon decay products to total lung dose varies from 25% to about 50%, depending on the uranium content in the ore being mined. Therefore, the risk per unit exposure of radon decay products, ERR/WLM, is likely overestimated by a factor of at least 2, possibly much more in high grade mines, in earlier years of mining when airborne ore dust was more abundant and if it were possible to take into account the doses accumulated in mines other than those subject to epidemiological studies. Table 10. Ration of the contribution of radon decay products and other sources to the dose to the lung in uranium mines. (H3 and H7 represent the equivalent lung dose calculated with wR = 3 and 7). ERn /E DRn /D H3, Rn /H3 H7, Rn /H7 low grade 0.56 0.48 0.57 0.61 highgrade 0.34 0.25 0.36 0.41 8 In uranium miners only the lung is the only organ at risk of radiogenic cancer . However, the effective dose to the lung and to other the organs are comparable and should induce approximately the same number of cancers [13, 14]. These cancers are not observed, in any cohort of uranium miners (Table 11). That observation begs the question of the effectiveness of combined low doses of gamma and alpha radiation to induce cancer at the levels encountered in uranium mines. Table 11. Comparison of the incidence of lung cancers and other cancers in 10 cohorts of uranium miners (Non-uranium Chinese tin miners excluded) . Collective exposure (WLM) 6.15E+06 Collective lung dose (Rn-D.P.) (Sv) 3.07E+04 Collective dose (other organs* ) (Sv) 3.07E+04 Expected (no. cases) 516 Lung cancers O/E = 3.22 (p<0.0001) Observed (no. cases) 1659 Observed (no. cases) 1179 Expected from ref. population (no. O/E = 1.01 1191 Other cancers cases) Predicted by ICRP risk coefficient Obs/Pred = 0.64 (0.59 - 0.69) 1605 (no. cases) p<0.0001) 3. CONCLUSIONS 3.1. Animal data At this stage of statistical analysis of the database, evidence of U-shaped dose responses is found for each type of radiation except for alpha emitters, for which the frequency of apparent U- shaped responses is not statistically significant. That evidence is not restricted to a single animal species, sex of the animal or type of cancer. Owing to the large range of experimental conditions (species, age at exposure, sex, type of radiation and of exposure, dose rates) the analysis of the database has required the development of new analytical techniques. The next steps include an evaluation of the quality of the available data, a review of the bio logical bases for U- based on original authors interpretation, a meta-analysis of observed reverse U-shapedness, and the publication of whole database. In the interpretation of the animal data presented in this paper, it is important to keep in mind that animal experiments on radiation carcinogenesis were designed to detect and measure a risk, not to examine whether exposure to ionizing radiation would induce detrimental or beneficial effects. Whether that specific objective resulted in missing information on decreased radiation carcinogenesis at low doses and dose rates remains to be investigated. 3.2. Human data A few epidemiological studies were conducted on radiation workers, with comparable reference populations. In these studies, no “health worker effect” can explain the statistically significant reduction immortality rates, from cancer or other causes, observed in occupationally exposed workers. Together, animal studies and human epidemiology suggest that, in some populations, beneficial effects of low-dose exposure to ionizing radiation are observed. They appear to reduce the risk of developing cancer or other diseases and to increases longevity. These observations are not new but the work presented here brings additional statistical evidenc e of the reality of these phenomena. 9 REFERENCES 1. International Radiobiology Archives. US DOE Report DOE/RL/96-72, (1996). 2. Gerber G. et al. International radiobiology archives of long-term animal studies: structure possible uses and potential extension. Radiat. Environ. Biophys. 38, 38-75 (1999). 3. P. Duport. A Database of Cancer Induction by Low-Dose Radiation in Mammals: Overview and Initial Observations. 2nd Conference of the World Council of Nuclear. Dublin (2001). 4. Maisin J.R. et al. Survival and diseases in C57 Bl mice exposed to X-rays or 3.1 MeV neutrons at an age of 7 or 21 days. Radiat. Res. 146, 453-460 (1996). 5. R.L. Ullrich, R. L. and J.B. Storer. Influence of γ irradiation on the development of neoplastic diseases in mice. Radiat. Res. 80, 303-316 (1979). 6. Berrington et al. 100 years of observation on British radiologists: mortality from cancer and other causes 1897-1997. Br. J. Radiol. 74, 507-519, (2001). 7. G. Matanoski. Health Effects of Low-Level Radiation in Shipyard Workers. US DOE Report, Contract DE-AC02-79EV10095, (1991). 8. BEIR VI. Health Effects of Exposure to Radon. National Research Council (USA), National Academy Press (1998). 9. Evans R.D. The effects of skelatally deposited Alpha-ray emitters in man. Brit. J. Radiol. 39, 881-895 (1967). 10. Andersson, M. and Strom, H.H. Cancer incidence among Danish Thorotrast patients. J. Natl. Cancer Inst. 4, 1318-1325 (1992). 11. Monchaux G., Morlier J-P., Altmeyer S., Debroche M., Morin M. Influence of exposure rate on lung cancer induction in rats exposed to 222 Rn D.P. Radiat. Res. 152, S137-S140 (1999). 12. Sanders C.L., McDonald, K.E., Mahafey, J.A. Lung tumor response to inhaled Pu and its implications for radiation protection. Health Phys. 55(2), 455 (1988). 13. P. Duport. Is radon risk overestimated? Neglected doses in the estimation of the risk of lung cancer in uranium underground mines. Radiat. Prot. Dosim. 98(3), 329-338 (2002). 14. International Commission on Radiological Protection. 1990 Recommendations of the ICRP. PergammonPress (1991). 15. Darby et al. Radon exposure and cancers other than lung in underground uranium miners: a collaborative study of 11 cohorts. J. Natl. Cancer Inst. 87, 378-384 (1995).