Fitting the Armitage-Doll model to radiation-exposed cohorts and implications for population cancer risks

The Armitage-Doll model of carcinogenesis is fitted to Japanese bomb survivors with the DS86 dosimetry and to three other radiation-exposed cohorts. The model is found to provide an adequate description of solid cancer incidence and also, to a lesser extent, of that of leukemia as a function of radiation dose when up to two radiation-affected stages are assumed. For non-leukemias the optimal model is one in which there are two radiation-affected stages separated by two additional stages. In the case of leukemia one radiation-affected stage or two adjacent stages provide suitable fits. There appear to be significant differences between the optimal models fitted to each cohort, although there is no heterogeneity within the Japanese data set by sex, by cancer type, or by age at exposure. Low-dose and low-dose-rate population risks for a population having the cancer and overall mortality rates of the current UK population are calculated on the basis of the optimal models fitted to the Japanese data to be about 8.3 x 10(-2) excess cancer deaths person-1 Sv-1, 10.1 x 10(-2) radiation-induced cancer deaths person-1 Sv-1, or 1.40 years of life lost person-1 Sv-1. Risks for a population having the mortality rates of the current Japanese population are about 6.5 x 10(-2) excess cancer deaths person-1 Sv-1, 7.8 x 10(-2) radiation-induced cancer deaths person-1 Sv-1, or 0.89 years of life lost person-1 Sv-1. It is a feature of the Armitage-Doll model, and other multistage models of carcinogenesis, that if radiation acts at more than one stage then (inverse) dose-rate effects may arise as a result of interactions between the effects of a protracted dose at the various radiation-affected stages. However, it is shown in this paper that these three measures of cancer risk in general display fairly slight dependence on administered dose in the range 0.001 to 1.0 Sv and on the length of the time over which the dose is administered in the range 1 to 100 years. Dose-rate effects resulting from the protraction of a radiation exposure over many years acting on (the same) cells at various stages of a multistep process of carcinogenesis are therefore expected to be slight. Dose-rate effects which have been observed in epidemiological studies and cellular radiobiology may thus find their explanation in other phenomena such as short-term intracellular repair.