| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Commentary |
1 Regional Radiation Protection Service, Addenbrookes Hospital, Hills Road, Cambridge CB2 2QQ, UK and 2 National Institute for Public Health and the Environment (RIVM), PO Box 1, 3720 BA Bilthoven, The Netherlands
The 20th L H Gray Conference held in 2002 took as its general title "Radiation Cancer Analysis and Low Dose Risk Estimation". The purpose of the meeting was to bring together experts from a broad spectrum of disciplines, including epidemiologists, modellers, statisticians and radiation biologists to discuss new insights in radiation-induced carcinogenesis and the implications for low dose radiation risk.
The findings of the Conference are also especially timely for the large number of staff in the healthcare professions who are concerned with the diagnostic uses of ionizing radiation. The Ionising Radiation (Medical Exposure) Regulations 2000 [1] and the extensive Guidance Notes to these regulations recently published [2] rely heavily on the linear-no-threshold (LNT) model of radiation induced cancer, especially with regard to the principles of Justification and Optimization. Therefore it is important to be aware of the latest findings, both for and against the model, and the degree of confidence that can be attached to it.
A major aim of the Conference was to look at the extent to which multistage cancer models can bridge the gap between mechanistic information such as in vitro dose responses and epidemiological information from which risks are estimated with an LNT approach. Four broad areas of research provided the majority of the papers; 1) cancer modelling, 2) basic mechanisms and bystander effects, 3) radon exposure and lung cancer risks, and 4) cancer after radiotherapy.
Modelling
The two stage clonal expansion model first proposed by Moolgavkar and Knudson [3] is now commonly used as the starting point for modelling radiation-induced and radiation driven carcinogenesis. The model is illustrated schematically in Figure 1
.
|
) and death rate (
) is not mutagenic. Radiation may have a role in stimulating cell growth at this stage (promotion) but this cannot be either confirmed or refuted by fitting the model to cancer data. Different approaches are taken when using the model to analyse data, with some groups using a strictly stochastic approach while others use a slightly less mathematically rigorous approach. The model does describe the age-dependent increase of many naturally occurring cancers as well as that of radiation-induced cancers and reveals an excess risk pattern which is close to, but not exactly the same as, relative risk. It also provides a better insight into the differences in cancer induction by chronic and acute exposures, elaborating on the simple linear–quadratic dose relationship. These three aspects are, for example, all discussed in the article by Leenhouts et al [4]. The model also offers an explanation of the threshold-like dose effect relationship found for bone cancers induced in the radium dial painters.
The model provides a useful research approach for estimating low-dose and low-dose rate radiation risks for specified exposures, in comparison with other models including the LNT concept. More biological information on the specific dose–response relationships for the model parameters is, however, needed before this model can be used with confidence to establish risks at dose levels where experimental verification is difficult if not impossible.
Basic mechanisms and bystander effects
A good example of a phenomenon that may increase the complexity of modelling is the bystander effect. This is defined as "a biological response in cells that do not themselves receive any energy deposition from ionizing radiation but respond to signals produced by cells that do". For a good review see Mothersill and Seymour [5].
Evidence for this effect has been increasing during the last decade from both microbeam as well as low dose broad beam experiments and three papers at the Conference examined the potential implications for low dose risk estimation. Mechanisms are not yet understood. For example how might the end point of cancer induction depend on a) the percentage of directly hit cells, b) the amount of energy deposited in directly hit cells, c) the spatial distribution of energy in spatially hit cells? Also what are the molecular weights/diffusion characteristics of the molecules carrying the signals? Are gap junctions and/or cell-to-cell communication mechanisms important? Finally, the effect has only been demonstrated to date in vitro and most convincingly with alpha particles. We do not know if this is a general effect with all ionizing radiation exposures or, at low doses, is restricted to high LET radiations (the experiments are technically much easier with alpha particles). We do not know if the broad beam and microbeam experiments result from the same mechanism, or whether the effect is significant for in vivo systems.
The bystander effect may have implications for extrapolation models of low dose radiation risk but it is not yet clear what they are. Arguments have been put forward for the effect either reducing the low dose radiation risk, due to extra cell killing, or increasing the risk, due to, for example, enhanced gene mutations and cell transformations in cells that would be killed at higher doses [6].
One Conference subject that should not be overlooked within the context of this "Commentary" relates to variation with photon energy of the relative biological effectiveness (RBE) of mutation induction and neoplastic transformation [7]. Experiments were performed in vitro on human fibroblasts and hybrid cells and showed that RBE values of mammography X-rays relative to 200 kVp X-rays or Co-60 gamma rays were approximately 4 and 8, respectively. The difference was attributed to the difference in the secondary electron spectrum of the different radiations with the 10–25 keV X-rays perhaps producing ion densities that are more likely to cause multiple double strand DNA breaks by single electrons.
Further work is required in this area and there is no hard evidence yet that this finding translates to in vivo but there may be potential implications for risk estimation in mammography.
Radon exposure and lung cancer risks
Most of the epidemiological studies looked at the risk of lung cancer. Case control studies are generally considered to be more accurate than ecological studies because they match individual cases with one or more individual controls but suffer from the limitation that a detailed exposure history has to be obtained. The strength of such studies is that they provide cancer risk estimates that cannot be derived from cellular studies.
Unfortunately, epidemiological studies cannot tell us if doses smaller than 10 mSv are carcinogenic or not. Using current excess relative risks of between 0.4 and 0.8 per Sv for solid cancer, if the natural incidence is 10% over 30 years and 1 Sv is taken as a doubling dose, at 10 mSv the total risk is only 10.1%. On purely statistical criteria, a cohort of 62 000 persons would have to be followed to have an 80% chance of finding an effect of 10 mSv at the 5% level. Taking the inevitable uncertainties in epidemiological studies also into account, it becomes clear that the cohort would have to be unrealistically large.
Several of the papers confirmed, if any confirmation were necessary, the strong link between smoking and cancer. Possible synergistic effects between radiation and smoking are unclear. Whereas the graph of incidence of lung cancer against radon concentration rises more steeply for smokers than for non-smokers, the excess relative risk is actually higher for non-smokers, although this is simply a consequence of the rather low risk for lung cancer in non-smokers.
It was possible to conclude that the absolute radon risk for non-smokers is real but small. Attempts to be more quantitative are confounded by poor information on smoking habits, potential passive smoking, the possibility that dust might be a promoter and other uncontrolled factors.
Cancer after radiotherapy
The problems associated with attempting to base risk estimates on patients being treated with radiotherapy have been well documented and have not been overcome. Nevertheless, some papers showed that carefully designed work can still provide useful information. In particular, increasing precision on the geometrical relationships between the irradiated tumour and adjacent anatomy and on the precise dose distribution help to identify radiotherapy-related parameters that influence the development of second malignancies. Dörr and Herman [8] reported that for both malignant and benign disease treated with radiotherapy, about 50% of the new tumours were within the margin region of the planning target volume which was defined as the volume from 2.5 cm inside to 5 cm outside the field margin proper. Long follow-up times are essential for such studies but the results may be important when using multiple field techniques or intensity modulated radiotherapy where the relevant volume is substantially increased.
The largest study in this category included 6841 patients from Sweden, Italy and France who had received I-131 radioiodine. 558 cases of second malignant neoplasms had been recorded but it is difficult to see how information on the administered activity could be converted into meaningful dosimetry.
A question of opinion
This was an interesting exercise in audience participation. A list of topically controversial questions had been prepared under three headings — epidemiology, multistage cancer modelling and risk — and participants were asked to express their opinion on each question by key-pad response. Note that the purpose was to stimulate discussion, not to determine the "true" answers. Surveys were made both at the start of the Conference and again at the conclusion to detect any significant change in opinion.
Space does not permit analysis of all 27 questions and responses, which the interested reader can find on http://www.rivm.nl/rca (click "Results" or "A question of opinion"). A selected cross-section that have particular relevance in the medical context are presented. Figures are given for the opinions of participants at the end of the Conference — although "before" and "after" figures were often very similar.
Epidemiology
74% of participants took the view that it was not always appropriate to assume a linear dose–effect relationship for the evaluation of data at low doses (<2 Gy). However, 79% believed there is not sufficient evidence of a threshold for cancer induction at low doses to justify a change in current radiation protection thinking. This response also effectively precludes radiation hormesis as a significant factor influencing radiation risk estimation at the present time.
Multistage cancer modelling
80% of participants believed that radiation at low dose does act as a cancer initiator and 78% believed it also acts on other mutational events in the cancer process. Opinion was much more divided on the role of low dose radiation as a promoter ("no" and "don't know" combined to 56%).
Risk
60% of participants believed that direct mutation induction had been demonstrated convincingly enough to be used as the basis for risk estimation extrapolations. The existence of bystander effects appears to have been proven unambiguously in some systems but 63% felt it would be premature to take this effect into consideration in risk assessments.
In 1995 Cohen [9] reported a strong negative correlation between lung cancer and radon exposure in USA counties and calculated a discrepancy with the positive correlation by the linear-no-threshold theory of more than 20 standard deviations. There has subsequently been extensive correspondence, especially in Health Physics and The Journal of Radiological Protection. 80% of the participants were of the opinion that the negative risk was not a radiation effect but the Conference was undecided on the need to reduce levels of indoor radon below 200 Bq.m-3 (25% yes, 52% no, 23% don't know).
There was strong support for age-dependent radiological risk factors being incorporated into radiological protection (96%). Recent figures suggested for the fatal cancer risk coefficient as a function of age at exposure are shown in Table 1.
|
If these data are combined with data on the distribution by age of patients undergoing medical radiological procedures (Table 2), it is clear that collective effective doses in diagnostic radiology seriously overestimate the collective risk if they are combined with a fixed risk factor of 7.3% Sv-1 for the risk to the whole population from fatal cancer, non-fatal cancer and hereditary effects [11].
|
No spectacular developments on the vexed question of low dose radiation risk were reported. For the foreseeable future cancer risk estimation for very low doses (10 mSv) will be based on informed risk estimates at moderate to high doses (0.2 Sv and higher).
Examination of available data leads to the conclusion that epidemiological studies will never be strong enough to resolve the threshold/no threshold debate unambiguously. Having said this, it can be concluded from re-examination of the data on childhood cancer that the risk following acute exposure in utero to doses of the order of 10 mSv is not zero.
In the context of radiological protection, upper credibility limits for low dose risk are more important than lower credibility limits. Therefore we should perhaps be more concerned about bystander effects than hormesis. The probability that a threshold exists has to reach almost one before the upper credibility limit (95%) on the excess relative risk per sievert begins to come down.
Modelling may be the way forward and may, ultimately, provide a more scientific approach to estimating risk. The two-mutation model has the merit of simplicity and has already been shown to be a more effective predictor of risk than linear interpolation. However, the development of good radiobiological models needs a deeper understanding of both the mechanisms of radiation action and the cellular changes that lead to malignancy. Biophysical studies strongly favour no threshold models since they suggest that double strand breaks in DNA are inducible by a single electron in a cell.
For practical radiation protection and legislation the LNT model remains the best pragmatic approach: a) erring on the side of caution; b) underpinning the ALARA principle; c) permitting dose summation.
For risk estimation in diagnostic radiology, a move towards age-related risk factors is welcomed and the news that the BEIR 7 report, due to be published shortly, will quantify the uncertainties associated with risk estimates is excellent. A chain is only as strong as its weakest link and increasingly sophisticated methods of measuring doses to patients may be of limited value if risk estimation is the dominant uncertainty.
The proceedings have been published as a Supplement to the Journal of Radiological Protection – Volume 22, Number 3A, September (2002).
Acknowledgments
The authors acknowledge with thanks useful discussions with Dr H P Leenhouts and Dr K H Chadwick.
Received for publication November 5, 2002. Revision received June 10, 2003. Accepted for publication July 17, 2003.
References
This article has been cited by other articles:
 |
|
 |
|
  P P Dendy Radiation risks in interventional radiology Br. J. Radiol., January 1, 2008; 81(961): 1 - 7. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. S. Storm, D. L. Miller, L. J. Hoover, J. D. Georgia, and T. Bivens Radiation Doses from Venous Access Procedures Radiology, March 1, 2006; 238(3): 1044 - 1050. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S Anoopkumar-Dukie, A McMahon, A Allshire, and T J Conere Further evidence for biological effects resulting from ionizing radiation doses in the diagnostic X-ray range Br. J. Radiol., April 1, 2005; 78(928): 335 - 337. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L E Feinendegen Evidence for beneficial low level radiation effects and radiation hormesis Br. J. Radiol., January 1, 2005; 78(925): 3 - 7. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
