by Dale L. Preston, Donald Pierce, and Michael V誥h, Department of Statistics, RERF

This article was originally published in RERF Update 4(4):5, 1992-93.

We do not find Tore Straume’s approach to risk evaluation appropriate nor his conclusions convincing. Our primary objection concerns the unquestioned application of the in-vitro results obtained by R. L. Dobson et al (Radiat Res 128:143-9, 1991) to the RERF chromosome aberration study. There are, of course, many pitfalls in extrapolating from experimental in-vitro results to those expected in human populations. Moreover, there is a great deal of pertinent experimental evidence that has been ignored in his argument.

Conclusions about cancer risks depend mainly on what is learned from the high-dose part of the data from RERF and elsewhere, combined with the most thoughtful use of external radiobiological knowledge about the relative biological effectiveness (RBE) of neutrons, dose rate effects, and the shape of the dose-response curve. The latter category of knowledge will not be affected by possible revisions of the RERF doses. There is little direct information about RBE in the RERF cohort data. To the small extent that there may be such information, the data would support smaller RBE values if the neutron doses were increased.

We also think that it is easy to misinterpret the implications of the Figure that accompanies Straume’s Update article. As discussed below, the modified Hiroshima neutron doses would be about 5% of the gamma doses. His results for the 1000- to 1200-m range, ie, a roughly equal contribution to risk from gamma and neutron, would then indicate only that the RBE is about 20 for doses at that range. His suggestion of a much higher RBE for the low-dose range is not novel, and, although this is an important issue, it has little bearing on risk estimates computed from the RERF cohort data, under either DS86 or the modified dosimetry.

It is more useful to see what effect the neutron changes he suggests would have on the RERF cancer risk estimates, as they are usually computed. The Table on this page shows the percentage decrease in cancer risk estimates if the Hiroshima neutron doses were to be increased as suggested by Straume et al (Health Phys 63:421-6, 1992). The first row results from an analysis in terms of total dose, whereas the results in the other rows were computed using dose equivalents. It is important to realize that linear cancer risk estimates from the RERF data depend on an assumed RBE largely through its value at high doses, around 1-1.5 Gy kerma. A large assumed low-dose RBE has little effect on linear dose-response analysis, both because of the unexposed group and because a large low-dose RBE results in little absolute change in total dose equivalent at low doses. This is the reason for use of the simple tentative dose equivalents in these calculations.

Straume et al (loc cit) originally presented tentative increases in Hiroshima neutrons as a function of slant distance. We were able to apply this relation directly rather than by using surrogate dose categories as in the Table of Straume’s Update article. The resulting increases are slightly larger than the factors (R) given for the dose categories of his Table. Our results pertain to risks for all cancers except leukemia. They are based on a standard model in which the excess relative risk is allowed to vary with sex and age at exposure, but is assumed to be constant in time.

It is surprising to many that the risk estimates in terms of total dose would only decrease by 3%. The reason for this can be understood by considering the nature of dose-response analysis along with information in Straume’s Table. As noted above, the primary effect on the dose-response slope arises from changes in total dose in the higher dose range, ie, at a distance of around 1000-1200 m. At that range, the current Hiroshima neutron estimates are about 1.5% of the gamma doses, and the tentative increase in neutrons is a factor of roughly 2.5-3. Thus, at that distance, the neutron component would increase from about 1.5% to about 4% of the gamma dose, resulting in an increase of about 2.5% (1.04/1.015) in the total dose.

Although it is important to remove as many systematic errors as possible in computing the cancer risk estimates, it should be realized that there are many uncertainties and factors that influence the risk estimates to about this extent. For example, RERF has recently found that adjustment for effects of random errors in the dose estimates, which has not previously been done in the major reports, should result in an increase in the risk estimates in the range of 10%-15%. Thus, changes of the magnitude indicated in the Table above should not, when taken in perspective, lead to substantial revision in assessments of the cancer risks of radiation.