Application of a state-vector model for radiation carcinogenesis to exposures of radon progeny in the lung: Test of the coherence between in vitro and in vivo models

The purpose of this research was to determine if a radiation carcinogenesis model capable of capturing known aspects of in vitro studies of cell transformation and death could adequately predict in vivo (rat) lung cancer incidence, after adjusting the model parameters for known biological differences and incorporating the dose rate distributions present in vivo but not in vitro. Constructing a more coherent analytic framework for linking in vitro cell transformation data and in vivo rat data has the potential to reduce some of the uncertainty in modeling lung carcinogenesis following exposure to radon progeny. Three forms of coherence exist when extrapolating from in vitro to in vivo levels of biological organization (Crawford-Brown and Hofmann 1996): ontological, formal and numerical. A completely coherent modeling framework requires that all three forms be identical at the two levels of organization, or that any differences be understood and incorporated into the model when extrapolating between the two settings.; The state-vector model used in the present research incorporated the biological mechanisms of DNA damage, DNA repair, chromosome aberration formation, cell differentiation, cell death, cell proliferation and intercellular communication. This model was first adjusted to correctly reproduce in vitro transformation frequency and cell killing data from C3H 10T1/2 mouse cells exposed to 150 keV/mum alpha-particles (Miller et al. 1995). The model was then adapted to incorporate dose rate distributions, in order to deal with the differences between in vitro situations (in which doses are uniformly distributed) and in vivo settings (in which doses are inhomogeneously distributed). Applying the model to in vivo rat data resulted in lung tumor incidence predictions that were compared against measured lung cancer data from rats exposed to radon progeny at the Battelle Pacific Northwest Laboratory (Cross et al. 1984; Cross 1987; 1988). Results of these comparisons show that the model is as yet unable to correctly reproduce in vivo dose-response curves. Therefore, the current version of the model is too problematic to justify its application for radiation protection purposes to human exposure data.