METHOD OF THE GLOBAL FITTING OF THE GENERALIZED RADIOGENIC RISK MODELS TO THE JAPANESE COHORT DATA

Research and Technical Center of Radiation-Chemical Safety and Hygiene (RTC RCSH), Moscow, Russia Generalized radiogenic risk models presented in ICRP Publication 103 have a number of attractive features. The models themselves and the corresponding models of background risks are generalized by a single function containing only three parameters specific to the localization of cancer, sex and risk indicator (absolute or relative; mortality or morbidity). However, there are several discrepancies in the values of model parameters for the individual cancer sites and in some predictions based on these models. The aim of this study was to clarify the possibility of refining the parameters of the ICRP models using the LSS cohort data. The method of calculation of the parameters of these models has been developed and verified based on the data on baseline mortality in the cohort and among entire Japanese population. The method consists in the averaging of the local estimates of the parameters calculated based on the values of the rate of the mortality in the ends of different segments of age at exposure and attained age within the entire examined area of change of these variables. The latest published mortality data in the cohort and mortality tables for Japan from the WHO database were used to calculate the mortality rate matrices. Data on the baseline mortality from solid cancers in LSS were in good agreement with statistically more reliable data for the entire Japanese population. Hence, the latter were used for verification of the method.According to the criteria of absolute and relative standard deviation, the ICRP models at whole better approximate baseline and radiogenic mortality from solid cancers in the cohort based on the values of parameters calculated using the proposed method compared to the values from the ICRP. The sets of parameter values for men and women, calculated from LSS data and for the Japanese population differ significantly, while in ICRP models they are almost the same.