Epidemiological studies have been carried out to determine the possible carcinogenic risk of doses lower than 100 mSv, and they have not been able to detect statistically significant risks even on large cohorts or populations. Therefore, these risks are at worse low since the highest limit of the confidence interval is relatively low. It is highly unlikely that putative carcinogenic risks could be estimated or even established for such doses through casecontrol studies or the follow-up of cohorts. Even for several hundred thousands of subjects, the power of such epidemiological studies would not be sufficient to demonstrate the existence of a very small excess in cancer incidence or mortality adding to the natural cancer incidence which, in non-irradiated populations, is already very high and fluctuates
according to lifestyle. Only comparisons between geographical regions with high and low natural irradiation and with similar living conditions could provide valuable information for this range of doses and dose rates. The results from the ongoing studies in Kerala (India) and China need to be carefully analyzed.

     Because of these epidemiological limitations, the only method for estimating the possible risks of low doses (< 100 mSv) is extrapolation from carcinogenic effects observed between 0.2 and 3 Sv. A linear no-threshold relationship (LNT) describes well the relation between the dose and the carcinogenic effect in this dose range where it could be tested. However, the use of this relationship to assess by extrapolation the risk of low and very low doses deserves great caution. Recent radiobiological data undermine the validity of estimations based on LNT in the range of doses lower than a few dozen mSv which leads to the questioning of the hypotheses on which LNT is implicitly based:

     1) constancy of the probability of mutation (per unit dose) whatever the dose or dose rate,
     2) independence of the carcinogenic process which after the initiation of a cell
evolves similarly whatever the number of lesions present in neighboring cells and the tissue.

     Indeed,
     1) progress in radiobiology has shown that a cell is not passively affected by the
accumulation of lesions induced by ionizing radiation. It reacts through at least three
mechanisms: a) by fighting against reactive oxygen species (ROS) generated by ionizing radiation and by any oxidative stress, b) by eliminating injured cells (mutated or unstable), through two mechanisms: i) apoptosis which can be initiated by doses as low as a few mSv, thus eliminating cells the genome of which has been damaged or misrepaired, ii) death of cells during mitosis when lesions have not been repaired. (Recent works suggest that there is a threshold of damage under which low doses and dose rates do not activate intracellular signalling and repair systems, a situation leading to cell death.) c) by stimulating or activating DNA repair systems following slightly higher doses of about ten mSv. Furthermore, intercellular communication
systems inform a cell about the presence of an insult in neighboring cells. Modern transcriptional analysis of cellular genes using microarray technology reveals that many genes are activated following doses much lower than those for which mutagenesis is observed. These methods have been a source of considerable progress by showing that depending on the dose and the dose rate not the same genes are transcribed.

     At doses of a few mSv (< 10 mSv), lesions are eliminated by disappearance of the cells; at slightly higher doses damaging a large number of cells (therefore capable of causing tissue lesions), repair systems are activated. They permit cell survival but may generate misrepairs and irreversible lesions. For low doses (< 100 mSv), the extent of mutagenic misrepairs is small but its relative importance, per unit dose, increases with the dose and dose rate. The duration of repair varies with the complexity of the damage and its amount. Several enzymatic systems are involved
and a high local density of DNA damage may lower their efficacy. At low dose rates the probability of misrepair is smaller. The modulation of the cell defense mechanisms according to the dose, dose rate, the type and number of lesions, the physiological condition of the cell, and the number of affected cells explains the large variations in radiosensitivity (variations in cell mortality or the probability of mutations per unit dose) depending on the dose and the dose rate that have been observed. The variations in cell defense mechanisms are also demonstrated by
several phenomena: initial cell hypersensitivity during irradiation, rapid variations in
radiosensitivity after short and intense irradiation at a very high dose rate, adaptive responses which cause a decrease in radiosensitivity of the cells during hours or days following a first low pre-conditionning dose of radiation, etc.

     2) Moreover, it was thought that radiocarcinogenesis was initiated by a lesion of the genome affecting at random a few specific targets (proto-oncogenes, suppressor genes, etc.). This relatively simple model, which provided a theoretical framework for the use of LNT, has been replaced by a more complex one including genetic and epigenetic lesions, and in which the relationship between the initiated cells and their microenvironment plays an essential role. This carcinogenic process is counteracted by effective defense mechanisms in the cell, tissue and the organism. With regard to tissue, the mechanisms which govern embryogenesis and direct tissue repair after injury appear to play also an important role in the control of cell proliferation. This is
particularly important when a transformed cell is surrounded by normal cells. These mechanisms could explain the lower efficacy of heterogeneous irradiation, i.e. local irradiations through a grid, as well as the absence of a carcinogenic effect in humans or experimental animals contaminated by small quantities of alpha-emitter radionuclides. The latter data suggest the existence of a threshold. This interaction between cells could also help to explain the differen ce in the probability of carcinogenesis according to the tissues and the dose, since the death of a large
number of cells disorganizes the tissue and favors the escape of initiated cells from tissue controls.

     All these data suggest that the lower effectiveness of low doses, or the existence of a practical threshold which could be related to either the failure of a very low doses to sufficiently activate cellular signalling and thereafter DNA repair mechanisms or to an association between apoptosis error-free repair and immunosurveillance.. However on the basis of our present knowledge, it is
not possible to define the threshold level (between 5 and 50 mSv?) or to provide the evidence for it. The stimulation of cell defense mechanisms, in particular to cope with reactive oxygen species. Indeed, a meta-analysis of experimental animal data shows that in 40% of these studies there is a decrease in the incidence of spontaneous cancers in animals after low doses. This observation has
been overlooked so far because the phenomenon was difficult to explain.

     These data show that it is not justified to use the linear no-threshold relationship to assess the carcinogenic risk of low doses observations made for doses from 0.2 to 5 Sv since for the same dose increment the biological effectiveness varies as a function of total dose and dose rate. The conclusion of this report is in fact in contradiction with those of other authors [43,118], which justify the use of LNT by the following arguments.

     1. for doses lower than 10 mGy, there is no interaction between the different physical events  initiated along the electron tracks through the DNA or the cell;
     2. the nature of lesions caused and the probability of error prone or error free repair and the elimination of damaged cells by cell death is neither influenced by the dose nor the dose rate;
     3. cancer is the direct and random consequence of a DNA lesion in a cell apt to divide and the probability of the initiated cell to give rise to cancer is not influenced by the damage in the neighbor cells and tissues;
     4. the LNT model correctly fits the dose-effect relationship for the induction of solid tumors in the Hiroshima and Nagasaki cohort;
     5. the carcinogenic effect of doses of the order of 10 mGy is proven for humans by results from in utero irradiation studies .

     The first argument concerns the initial physico-chemical events which are proportional to dose; however, the nature and efficiency of cellular defense reactions that are activated vary with dose and dose rate. The second argument is contradicted by recent radiobiological studies considered in the present report. The third argument does not take into account recent findings on the complexity of the carcinogenic process and the particular role of intercellular relationships and the stroma.. Regarding the fourth argument, it can be noted that besides LNT, other types of doseeffect relationships are also compatible with data concerning solid tumors in atom bomb survivors, and can also satisfactorily fit epidemiological data that are incompatible with the LNT concept, notably the incidence of leukemia in these same A-bomb survivors. Furthermore, taking into account the latest available data, the dose-effect relationship for solid tumors in Hiroshima-Nagasaki survivors is not linear but curvilinear between 0 and 2 Sv. Moreover, even if the doseeffect
relationship were demonstrated to be linear for solid tumors between, for example, between 50 mSv and 3 Sv, a generalization would not be possible because of experimental and clinical data show that the dose effect relationship considerably varies according to type of tumor and age of individuals at the time of irradiation. The global annd empirical relationship observed for solid tumors corresponds to the sum of relationships which can be quite different according to the type of cancer, for example, some being linear or quadratic, with or without threshold.

     Finally, with regard to in utero irradiation, whatever the value of the Oxford study, some inconsistencies between the availbable data sets call for great caution before concluding the existence of a causal relationship from data showing simply an association. Furthermore, it is highly questionable to extrapolate from the fetus to the child and adult, particularly, since the developmental state, cellular interactions and immunological control systems are very different.
 
 
 

      In conclusion, this report raises doubts on the validity of using LNT for evaluating the carcinogenic risk of low doses (< 100 mSv) and even more for very low doses (< 10 mSv). The LNT concept can be a useful pragmatic tool for assessing rules in radioprotection for doses above 10 mSv; however since it is not based on biological concepts of our current knowledge, it should not be used without precaution for assessing by extrapolation the risks associated with low and even more so, with very low doses (< 10 mSv), especially for benefit-risk assessments
imposed on radiologists by the European directive 97-43. The biological mechanisms are different for doses lower than a few dozen mSv and for higher doses. The eventual risks in the dose range of radiological examinations (0.1 to 5 mSv, up to 20mSv for some examinations) must be estimated taking into account radiobiological and experimental data. An empirical relationship which has been just validated for doses higher than 200 mSv may lead to an overestimation of risks (associated with doses one hundred fold lower), and this overestimation could discourage patients from undergoing useful examinations and introduce a bias in radioprotection measures against very low doses (< 10 mSv).

     Decision makers confronted with problems of radioactive waste or risk of contamination, should re-examine the methodology used for the evaluation of risks associated with very low doses and with doses delivered at a very low dose rate. This report confirms the inappropriateness of the collective dose concept to evaluate population irradiation risks.