Register for Updates | Search | Contacts | Site Map | Member Login


View Comment

Submitted by Prof Dr Dietrich Harder, Medical Physics and Biophysics
   Commenting as an individual
Document Health risks attributable to radiation
Comments on document 12/421/04 “Low-dose Extrapolation of Radiation-Related Cancer Risk” (Committee 1 Task Group Report) and on the related point 4.2.1 “Risk of cancer” in the “Draft for consultation” of the 2005 ICRP recommendations

To the chairman, members and secretary of the ICRP.

Dear colleges, in my understanding of the never-ceasing obligations as a former member of the ICRU and former chairman and member of the German Radiation Protection Commission I want to submit some critical remarks concerning ICRP’s approach towards the problem of extrapolating radiation-related cancer risk to low doses. Let me first say that I know ICRP’s commendable history in dealing with this fundamental task, e.g. from my earlier work as chairman of the “risk committee” of the German Radiation Protection Commission. I am convinced that ICRP should profit from collegial suggestions in drafting the new ICRP report on the basis of the preparation work already done in its committees. Please be assured that my criticism is coming from an old friend, although I will recommend a reconsideration of ideas.

My thesis:
At present, the linear dose-risk relationship without threshold can not be regarded as proven by radiobiological theory. For the time being, the ICRP might consider to qualify the LNT model as what it has been before: a simple fitting function, applicable for practical purposes of radiation protection, but not able to predict the numbers of cancer cases in the low-dose region on the basis of an established theory.

In publications in well-known journals as well as in official announcements the linear non-threshold model (“model” in the sense of a curve fitting function) of cancer risk versus dose is increasingly understood to represent a proven law of nature or at least a promising scientific hypothesis. On this conceptual background epidemiologists tend to claim that their study results are “compatible with the LNT hypothesis”, and they do this even at doses and dose rates as low as those of the average natural radiation exposure. By application of the linear dose-risk relationship without threshold, expectation values of thousands of cancer cases per country and year are then calculated. For instance 20000 cases of lung cancer per year in Europe have recently been attributed to radon in dwellings, with no hesitation in view of the fact that this result was obtained under application of the linear fitting function down to dose zero. The average natural radiation exposure, whose role in evolution, adaptive processes and homeostasis has long been regarded as not fully understood and to be clarified in further research, was thereby readily declared as carcinogenic, and thousands of cancer cases were simply calculated on the basis of linear curve fitting.

When you talk to the leading epidemiologists, addressing the large confidence intervals of the assessed risk data at low doses, they of course admit that - within certain bounds - their results are also compatible with dose-risk relationships with thresholds or with an upward curved initial part. In fact non-linear initial parts of the fitting functions have now been found in the LLS data even for solid tumours (Walsh, Rühm, Kellerer, Radiat. Res. 163, 477, 2005, and Preston, Pierce, Shimizu, Radiat. Res. 163, 477, 2005). Also for lung cancer risk by radon in dwellings, the existence of a threshold somewhere in the region below 150 Bq/m³, i.e. still above the radon concentration in many of the dwellings, has not been excluded (Darby et al. BMJ 330, 223-227, 2005, unabbreviated internet version of Dec. 21, 2004, see also statement of the German Radiation Protection Commission of April 22, 2005). A number of other human data showing dose thresholds or upward curved initial parts of the dose-risk relationship, e.g. for radium dial painters and for thorotrast patients, have been compiled in the well-known statement of the French Academy of Sciences of March 30, 2005. Moreover, a large number of dose thresholds for tumorigenesis observed in animal experiments has been compiled by Tanooka et al. (Int. J. Radiat. Biol. 77, 541-551, 2001). This large amount of evidence - not at the margin, but in the centre of our knowledge of radiogenic cancer risk - goes far beyond what has been called “recognized exceptions” in § 101 of the “Draft for Consultation” of the 2005 ICRP recommendations.

But admittedly, in the dose and dose rate range of the average natural radiation exposure, the limited accuracy of the empirical risk versus dose data does not permit any distinction between a dose-risk relationship of the LNT type or with a nonlinear shape. That is the origin of the question of the “Low-dose Extrapolation of Radiation-Related Cancer Risk” dealt with in ICRP document 12/421/04. The approach chosen by the ICRP to perform this extrapolation is radiobiological theory, and this approach was already started in ICRP 60. When you look for the core of the relevant statements, particularly of § 100, line 7: “no threshold for stochastic effects” and of § B61, line 2: “…do not support the idea of a threshold”, you finally identify this theory in § B48, line 1: “It is assumed that there is no threshold for the induction of the molecular change at specific DNA sites involved in the initial events that result in malignant transformation and ultimately cancer…” Although the subsequent steps of the multistage process of cancer development are briefly mentioned in this context, the formulation of § B48 leaves no doubt that the non-threshold concept of the ICRP is theoretically based on the assumption that the target for the carcinogenic action of radiation is the initiation phase of the multistage process of carcinogenesis. There is no discussion of the possible alternative idea that the target or targets of radiation carcinogenesis could be situated more downstream the chain of events, e.g. could consist in additional somatic mutation steps of the promotion phase or in an impairment of the immune defence against clonal expansion of precancerous cell clones. And even chapters 3 and 4 of the new 211-page ICRP document 12/421/04, which is entirely devoted to the question of the low-dose extrapolation of radiation-related cancer risk, are merely dealing with the initiation phase of radiation carcinogenesis. Neither the possible impairment of the immune defence by radiation nor even the immune system as such have been mentioned a single time in the whole document.

The ICRP has already received the well-founded protest of the French Academy of Sciences of March 30, 2005, against this attempt of a biologically quite narrow, too much simplified theoretical proof of the validity of the LNT hypothesis. It has been pointed out that the complexity of the multistage process of carcinogenesis also permits other extrapolations to low doses, and the assumed linear relationship has only been acknowledged as what it was before - a simple fitting formula. In fact, there are strong hints that, especially at low doses, the essential role of radiation is not in the initiation phase. For instance, lines of evidence come from the astounding effect that in many cases the radiogenic cancer risk is approximately of the multiplicative type, i.e. appears as a dose-dependent factor on top of the baseline risk. Almost the same dose-dependent multiplicator is valid even when the baseline risk is largely varying between population groups, such as the risk of breast cancer between Japan and USA, the risk of lung cancer between smokers and non-smokers, and the risk of leukemia between groups of different age or between groups with and without medicinal immune suppression. In recognition of this phenomenon, Nakamura (Radiat. Res. 163, 258-265, 2005) has proposed that the target point of ionizing radiation in the multistage process leading to leukemia may be the promotion phase where the radiation action provides the final somatic mutation steps, whereas the earlier mutation steps are the same as for the baseline risk. The same idea has been proposed, e.g. for lung cancer, in a mathematical model of radiation carcinogenesis (Heidenreich and Paretzke, Radiat. Res. 156, 678-681, 2001). Another possible target for radiation to interfere with the multistep process of carcinogenesis is the radiation-induced impairment of the immune defence against the expansion of precancerous cell clones. The existence of this mode of action can be predicted by extrapolation from the established mechanism of UVB-induced immune depression, leading to skin cancer. It can also be expected in view of the well-known occupation of the immature dendritic cells with the phagocytosis of radiation-induced apoptotic or necrotic cells (D. Harder, in “Radon as medicine”, ISBN 3-8300-1768-5), which will weaken their function in the framework of immunosurveillance. These proposed alternative mechanisms have the potential of predicting non-linear dose-risk-relationships.

In conclusion, the evidence presently accumulated and roughly outlined above does no longer permit the ICRP to state - as has been done in § 101 of the “Draft for consultation” of the 2005 ICRP recommendations - ”that the weight of evidence on the fundamental cellular processes supports the view that in the low dose range up to a few tens of mSv, it is scientifically reasonable to assume that in general and for practical purposes cancer risk will rise in direct proportion to the absorbed dose in organs and tissues.” Under consideration of the complexity of the multistage process of carcinogenesis, this has now become a weakened, probably untenable position, at least until further clarification of the alternatives.

A well-considered formulation how to deal, in this situation, with the LNT model has been proposed by Trott and Rosemann (Radiat. Environ. Biophys. 39, 79 – 87, 2000), and at the present time this proposal might indicate to the ICRP a useful and recommendable way to go: “The linear non-threshold dose response hypothesis may be used in radiation protection planning as a simple, convenient method to optimise procedures and regulations, but should not be mistaken as a stringent scientific conclusion directly derived from the present state of knowledge of the processes involved in radiation carcinogenesis.” Essentially the same proposal has been made by the French Academy of Sciences in their statement of March 30, 2005. In the practice of radiation protection, this formulation will not change the useful application of the LNT model for general and practical purposes, but will avoid the hitherto unjustified claim that the LNT model is the expression of a proven theory or known law of nature holding even in the dose range of the average natural radiation exposure. It will emphasize the task of finding in the future the true target point or points of ionising radiation in the multistage process of carcinogenesis, and it will avert any unwarranted constraints in understanding the role of the natural radiation exposure of man in the course of the evolution.

With all good wishes for the endeavours of the ICRP and with best personal regards,

Dietrich Harder, Göttingen.