Low-dose Extrapolation of Radiation-Related Cancer Risk

Draft document: Low-dose Extrapolation of Radiation-Related Cancer Risk
Submitted by Richard WAKEFORD, BNFL
Commenting as an individual

I enjoyed reading this draft report on the risk of cancer resulting from low doses of ionising radiation and I think that some good points are made well. I set out below some matters for the consideration of the Task Group. 1) I gained the impression when reading the draft report that it was aimed more at the notion that the dose-response may have a threshold than at the alternative view that risks from low doses have been underestimated by the currently adopted linear no-threshold model, perhaps seriously. Although the threshold argument does deserve close attention, this should not be at the expense of neglecting the idea that the current models underlying radiological protection underestimate the risks of low-level exposure. It is important to appreciate that the linear no-threshold model is not conservative in its predictions of risk against all alternatives. For example, Brenner et al. (Radiat Res 2001; 155: 397-401) have argued, based upon experimental evidence for the bystander effect, that the dose-response for cancer may turn upwards at low doses (the bystander effect being saturated at moderate and high doses), so that a linear model would underestimate the risk, possibly substantially. Little and Wakeford (Radiat Res 2001; 156: 695-699) have demonstrated from epidemiological evidence of excess lung cancer risk following exposure to radon that such an effect would be small (at least, under these circumstances), and Brenner and Sachs (Health Phys 2003; 85: 103-108) appear to have accepted this point (see also, Little, J Radiol Prot 2004; 24: 243-255). Other epidemiological evidence is available to place limits on the degree to which a linear dose-response model might underestimate the risk of radiation-induced cancer. Wakeford and Little (Int J Radiat Biol 2003; 79: 293-309) have noted that a review of the evidence for childhood cancer following intrauterine exposure to radiation indicates that the risk at low doses is unlikely to have been grossly underestimated by a linear model, although various uncertainties mean that stronger inferences cannot be made. The large radiation worker studies (e.g. Cardis et al., Radiat Res 1995; 142: 117-132, and Muirhead et al., J Radiol Prot 1999; 19: 3-26) also place limits on the level of risk resulting from low dose-rate exposures. 2) A related topic is the variation in individual sensitivity to radiation-induced cancer and radiosensitive sub-groups. Such sub-groups (which are known to exist to some small degree) could, if of a reasonable size and sensitivity, produce a bimodal dose-response, the steeper slope at low doses being influenced by the radiosensitive sub-group. There is some evidence that such sub-groups might exist for some cancers (e.g. Ohtaki et al., Radiat Res 2004; 161: 373-379, and Nakamura, Radiat Res 2005; 163: 258-265). I would have thought that this possibility needed to be addressed to some extent in the report. 3) I am surprised that no reference is made in the draft report to the work of Cohen (e.g. Health Phys 1995; 68: 157-174) on lung cancer following domestic exposure to radon, which is likely to attract criticism. I was impressed by the study of Puskin (Health Phys 2003; 84: 526-532), which, it seems to me, effectively deals with any threshold or hormesis inference drawn from the ecological studies of Cohen. The evidence for lung cancer following residential exposure to radon is now complemented by Darby et al. (BMJ 2005; 330: 223-228) and Krewski et al. (Epidemiology 2005; 16: 137-145), and shows an excess risk of lung cancer down to reasonably low levels of radon exposure. 4) The effect of radioactive fallout from atmospheric nuclear weapons testing, especially upon the risk of childhood leukaemia, has been studied. Darby et al. (BMJ 1992; 304: 1005-1009) conducted a geographical correlation study of childhood leukaemia in the Nordic countries following the peak of atmospheric nuclear weapons testing in the late 1950s and early 1960s. They found a marginally statistically significant rise in leukaemia among young children of a magnitude consistent with that predicted by standard risk models. Stevens et al. (JAMA 1990; 264: 585-591) carried out a case-control study of leukaemia in southwest Utah, a region that had experienced fallout from the Nevada Test Site. A significant excess risk of leukaemia among those exposed as children during the period of peak fallout was found, at a level compatible with the predictions of standard models. These studies demonstrate that it is likely that an excess risk of childhood leukaemia follows low dose, low dose-rate exposure to radiation, and that the magnitude of this risk is comparable with conventional radiation risk models. 5) Note that the statistical association between childhood leukaemia and the dose of external radiation received by fathers while employed as radiation workers prior to the child's conception relates to the West Cumbrian village of Seascale, adjacent to the Sellafield nuclear complex (see Page 22). The evidence that this association has not been found elsewhere has been reviewed by Wakeford (J Radiol Prot 2002; 22: 191-194), and the isolation of the association originally reported by Gardner et al. (BMJ 1990; 300: 423-429) probably merits further mention, since the overall epidemiological evidence now renders a causal explanation most unlikely. 6) Not much is said about the epidemiological evidence for a threshold, which could lead to criticism. I am thinking, in particular, about bone cancer after exposure to 226Ra (the radium dial painters), for which there is good evidence for an excess risk at high doses, but little evidence for an excess risk at low bone doses. This is used by some (e.g. Peckover and Priest, Nuclear Energy 2004; 43: 145-152) to argue that, at least for some cancer types, the dose-response is significantly sub-linear, and possibly has a threshold. There is also evidence for some other cancer types (e.g. non-melanoma skin cancer) that an excess risk exists at moderate and high doses, but not at low doses. The report should address such epidemiological evidence that the linear no-threshold dose-response may significantly over-estimate the risk of cancer, at least for some cancer types. 7) Little is said within the draft report about the latest studies of the Mayak workforce. These are impressive studies (although there are still issues to be sorted out, such as the plutonium dosimetry) that show clear and large excess risks related to protracted exposure to external radiation and plutonium. Of particular interest is the latest paper by Gilbert et al. (Radiat Res 2004: 162: 505-516) concerning lung cancer and plutonium exposure. The dose-response is linear with a significant excess risk apparent in the lowest plutonium dose group. Further, the plutonium-related lung cancer risk coefficient (Sv-1) compares well with the equivalent coefficient obtained from the Japanese atomic bomb survivors, indicating that the radiation weighting factor for plutonium of 20 is not too far adrift (although the uncertainties in the plutonium dosimetry must be borne in mind). 8) A comparison between exposure to internal emitters and exposure to external radiation has recently been made by Harrison and Muirhead (Int J Radiat Biol 2003; 79: 1-13), and the report could usefully refer to this review. 9) One point that I don't think that the draft report currently addresses is the possibility of ascertainment biases in the LSS mortality data, as opposed to the LSS incidence data. I am aware that much of the report makes use of the incidence data, but there are references to the mortality data, and these data show a somewhat unusual pattern for solid tumours at low to moderate doses. Placing a limit on the threshold using the LSS mortality data does produce a lower limit, but these data can also be used to point to a supra-linear dose-response, which is not present in the LSS incidence data. I think that the report should make reference to this problem, and caution against making too much of the LSS mortality data at low doses.