Modeling of cell inactivation and carcinogenesis in the atomic bomb survivors with applications to the mortality from all solid, stomach and liver cancer

Abstract

The two-stage clonal expansion (TSCE) model of carcinogenesis has been applied to cancer mortality data from the atomic bomb survivors, to examine the possible influence of radiation-induced cell inactivation on excess relative risk (ERR) and excess absolute risk (EAR) estimates. Cell survival curve forms being either conventional or allowing for low-dose hypersensitivity (LDH) were investigated. Quality-of-fit tests for non-nested models were used in comparisons with the types of empirical risk models applied at the Radiation Effects Research Foundation (RERF) in Hiroshima. In general the TSCE model was found to represent the data more economically (i.e., with fewer parameters for a similarly good description of the data) than the empirical risk model. However, the data are not strong enough to give a clear preference to one of the very different model types used. Central ERR and EAR estimates (at 1 Sv, for age at exposure 30 and age attained 70) from TSCE and empirical models were in good agreement with each other and with previously published estimates. However, the TSCE models including radiation-induced cell inactivation resulted in a lower estimate of the relative risk at young ages at exposure (0-15 years) than the empirical model. Also the TSCE model allowing for radiation-induced cell inactivation with a conventional cell survival curve resulted at 0.2 Sv in significantly lower risk estimates than the model with LDH. These model differences have been used here to suggest risk estimates which include model uncertainty as well as the usual statistical uncertainty. Model uncertainties were small for central estimates and larger for other values of the variables. Applying the proposed method to excess risk for all solid cancer at 1 Sv, age at exposure 10 and age attained 70, results in total uncertainty ranges that are wider than the pure statistical uncertainty range by about 30% for both ERR and EAR.