How is the risk of radiation-induced cancer influenced by background risk factors? Invited commentary on “A method for determining weights for excess relative risk and excess absolute risk when applied in the calculation of lifetime risk of cancer from radiation exposure” by Walsh and Schneider (2012)

Abstract

The study of the Japanese survivors of the atomic bombings of Hiroshima and Nagasaki in 1945 remains the primary (but certainly not the only) source of epidemiological evidence on the health effects consequent to exposure to ionizing radiation. Recently, expert groups, such as the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR 2008), have used the latest cancer incidence and mortality data from the Life Span Study (LSS) of the Japanese atomic bomb survivors (Preston et al. 2004, 2007) to develop risk models to describe how the radiation-induced excess risk of a number of different types of cancer varies with the tissue-specific dose of radiation received during the bombings, and also how the risk is modified by factors such as sex, age at exposure, and time since exposure (Wakeford 2012a, b). These risk models are expressed in terms of the excess relative risk (ERR, the proportional increase in risk in comparison with the background absolute risk in the absence of the exposure) or the excess absolute risk (EAR, the additional risk beyond the background risk) produced by exposure to ionizing radiation during the atomic bombings. In their paper, Walsh and Schneider (2012) suggest a new method for assessing the weights to be attached to the ERR and EAR model predictions when using a mixture of the two models to describe the radiationinduced risk—although the use of the terms ERR and EAR in this context is slightly misleading, as Walsh and Schneider (2012), with many others [US National Academy of Sciences 2006; International Commission on Radiological Protection (ICRP) 2007; Little et al. 2008; UNSCEAR 2008], have developed models of relative and absolute excess risk with adjustment for attained age, age at exposure, and other risk-modifying variables. The method is based on the established statistical technique of multi-model inference (MMI) (Burnham and Anderson 1998; Claeskens and Hjort 2008) previously used by the authors in assessing radiation risk (Schollnberger et al. 2012; Walsh and Kaiser 2011), and also used in many other contexts. Although not explicitly Bayesian, MMI is somewhat related to Bayesian model averaging and related Bayesian techniques (Wang et al. 2012); these Bayesian methods have the advantage of assessing the parameter uncertainty distribution more thoroughly than MMI, albeit at somewhat greater computational cost. In particular, we judge that the method of Walsh and Schneider (2012) may not adequately assess the uncertainties in the ERR/EAR weights and other model parameters, which Bayesian techniques can better address (Wang et al. 2012). The mixing of ERR and EAR models is an important subject because the radiation-induced risk of a particular type of cancer will be in addition to the background risk of that cancer, and if (and to what extent) radiation interacts with those factors that produce the background risk, then this will influence the magnitude of the excess risk attributable to radiation exposure. By investigating the mixture of ERR and EAR models that provides the optimal description of the radiation-induced risk of a particular cancer type in the Japanese atomic bomb survivors, Walsh M. P. Little (&) Radiation Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, DHHS, NIH, Rockville, MD 20852-7238, USA e-mail: mark.little@nih.gov