The ICRP draft report "Low-dose Extrapolation of Radiation-Related Cancer Risk" presents a well-researched discussion of many aspects of its subject but it is not comprehensive. Specifically, it has three important omissions, as follows: 1. the lack of adequate consideration of chronic exposures to background radiation, 2. the lack of adequate discussion of occupational exposures, and 3. the lack of proper consideration of the evidence for radiation hormesis. A major shortcoming is the absence of any discussion of epidemiology related to background radiation (except for some aspects of the discussion related to radon). This absence of discussion may be because there is no epidemiological evidence of risks from background radiation, but it is for exactly this reason that discussion should be included. A valid risk model should be applicable to natural background radiation, which varies around the world from <1 mSv/y to >100 mSv/y. The treatment of statistical significance in the ICRP draft report suggests that a study population of over 600,000 would be needed for a one-off dose of 10mSv followed by observation over a 30-year period. However, the position is likely to be very different for a continuing dose rate of 10mSv/y over 30 years. This will not be known unless epidemiology is applied. Most of the world’s population is exposed to background radiation levels in the range 1 to 10 mSv/y and there are substantial groups exposed up to at least 50 mSv/y. Applying the ICRP-recommended risk coefficient suggests that radiation-caused cancer rates would correspondingly range from 5 to more than 50 deaths per 100,000 per year, if the ICRP recommendation is correct. A typical figure for cancer deaths from all causes is about 150-200 per 100,000 per year. Hence, a difference of 45 cancer deaths per 100,000 per year due to radiation exposure would surely be discernable by epidemiology if it really happens. At present, there is no consistent evidence of a positive correlation between natural background radiation and the incidence of cancer. If anything, the cancer death rate appears to reduce with increase in background dose rate. For references and further discussion of this aspect, see: Higson, D.J. (2002), Resolving the Controversy on Risks from Low Levels of Radiation. The Nuclear Engineer, 43(5), September/October 2002, 132-137 As a related issue, it is difficult to understand how a DDREF of only 2 can be supported when there is no clear evidence of harm from lifetime doses up to at least a few thousand mSv from chronic exposures, e.g. due to natural background radiation. Another shortcoming of the ICRP draft report is its heavy reliance on the Cardis et al (1995) study, in relation to risks from occupational exposures. To quote from this study, it actually found that there is “no evidence of an association between radiation dose and mortality from all causes or from all cancers”, for individual doses up to 100 mSv. It did indeed find a slightly positive association between dose and the incidence of leukaemia, but leukaemia was quite a small part of the total. And, of course, the indication of a positive association for leukaemia must have been balanced by negative associations for other cancers, for there to be no association for the total. In fact, the purpose of the study was to look only for significant dose-related increases in mortality. Tests of statistical significance were applied only to the increases. Data showing significant dose-related decreases were, in effect, ignored. Also, it is not entirely clear whether the data from the US Nuclear Shipyard Workers Study, which indicated substantial radiation hormesis, was fully integrated into Cardis et al's study. Other studies since 1995, including a study of the Lucas Heights’ workforce, in Australia, have found results that are consistent with the action of hormesis at occupational exposure levels – see: Higson, D.J., Healthy Radiation Workers. “Radiation Protection in Australasia” Journal of the Australasian Radiation Protection Society, 19 (1), 2002. (This paper was also presented at the European IRPA Congress in Florence, Italy, October 2002.) The application of radiation hormesis, if it occurs, would be a matter for the medical profession, not for the radiation protection profession. However, the existence of radiation hormesis would invalidate the linear no-threshold (LNT) assumption. This would not be just a question of the risk being somewhere between zero and the value estimated by LNT (i.e. LNT being "conservative") but that low level radiation would actually be beneficial – so that the application of LNT and ALARA would actually increase risk. Nevertheless, Section 2.5 of the ICRP draft report does not include any discussion of the possibility of benefits from low level radiation. In 1990, the ICRP (in its Publication 60) conceded that hormesis might exist but said that “the available data on hormesis are not sufficient to take them into account in radiological protection”. With the publication of a great deal of evidence on hormesis since 1990, the converse is now true, viz: that the ICRP would need to be very confident that radiation hormesis does not occur if it is going to recommend the assumption of LNT. References to relevant information can be accessed through the following publications: Higson, D.J. (2002), Resolving the Controversy on Risks from Low Levels of Radiation. The Nuclear Engineer, 43(5), September/October 2002, 132-137 Higson, D. J. (2004), The Bell Should Toll for the Linear No-threshold Model. Journal of Radiological Protection, 24(3), September 2004, 315-319 Hormesis is mentioned twice in the ICRP draft report. The first time, on p.33, suggests that observations of radiation hormesis may be purely chance occurrences and may have been selectively presented to give the appearance of a substantial body of evidence. If this is to be the basis of ICRP's view of hormesis, it really does need to be substantiated. Otherwise, it appears to be an unwarranted aspersion on many highly reputable scientists, including: · Professor Ed Calabrese of the University of Massachusetts/Amherst, who has amassed and reviewed a huge number of literature references to hormesis; and · Dr Ron Mitchel of AECL's Chalk River Laboratories in Canada, who has published many papers purporting to demonstrate radiation hormesis by research on animals and by theoretical analyses. The second mention of hormesis in the ICRP draft report (on pages171-172) says that "estimation [of an increase in cancer rate] requires information on .... how to extrapolate risk from high to low doses and from high to low dose rates, including .... departures from the LNT hypothesis such as hormesis ....". There can be no disagreement that this is a requirement, but the ICRP draft report does not suggest how it is to be met. Currently, it must therefore be concluded that extrapolation (to dose rates less than about 100 mSv/y) is not justified. Indeed, there appears to be positive evidence against extrapolation.