NCRP Report 181, Evaluation of the Relative Effectiveness of Low-energy Photons and Electrons in Inducing Cancer in Humans: A Critique and Alternative Analysis.

A recent report from the National Council on Radiation Protection and Measurements presents an evaluation of the effectiveness of low-energy photons and electrons, relative to higher-energy photons, in inducing cancer in humans. The objective of that evaluation was to develop subjective probability distributions of an uncertain quantity, denoted by ρ, to represent ranges of credible values of the effectiveness of five groups of low-energy radiations (L): photons at about 1.5 keV; 15 to 30 keV photons; 40 to 60 keV photons; >60 to 150 keV photons; and electrons from beta decay of tritium (H). Probability distributions of ρL for all low-energy groups were derived based on an evaluation of uncertainties in data on biological effectiveness from five areas of research and use of an elicitation process and decomposition method to combine probability distributions to represent those uncertainties. In this paper, we argue that uncertainties in ρLs for all low-energy groups are too small compared with uncertainties in biological effectiveness from the different areas of research, especially that upper confidence limits of all ρLs are too low. These deflations of uncertainty in all ρLs apparently are due, at least in part, to an invalid assumption in the decomposition method that probability distributions of biological effectiveness from the different areas of research are representations of random uncertainty that arises from repeated measurements of the same quantity under the same conditions using well-calibrated instruments. However, those distributions essentially are representations of systematic uncertainty in different estimates of biological effectiveness from each area of research, which means that a deflation of uncertainty in ρLs is not a credible result. We then use the same probability distributions of biological effectiveness from the different areas of research in an alternative analysis to derive wider probability distributions of ρL that we believe provide a better representation of the state of knowledge of the effectiveness of low-energy photons and electrons in inducing cancer in humans. Our analysis is based on the notion that each probability distribution of biological effectiveness from an area of research represents a distinctly different model of a ρL and use of the concept of model averaging to combine those distributions.