estimate its values for a variety of in vitro end points that are considered to be relevant to cancer induction and to be adequately quantifiable and then to define from these data sets a single value that is judged to be applicable to human cancers overall. The use of in vitro oncogenic-transformation data as a basis for risk estimates for more complex end points, such as carcinogenesis in general in humans, has been discussed elsewhere (ICRU 1986). Essentially, the rationale, other than the pragmatic issue of quantifiability, is that the radiation weighting factor is used for predicting only relative risks (compared with risks associated with gamma rays or x rays) of one kind of radiation relative to another, rather than absolute risks. However, many data on in vitro effects or carcinogenesis in animals show that the RBEs for the same kind of radiation depend substantially on the biologic system and cancer type under study. The RBE for induction of lung-cancer by radon-progeny alpha particles remains uncertain.
On the basis of in vitro data on the C3H10T1/2 oncogenic-transformation system of Brenner and colleagues (1995) and data on the induction of micronuclei in rat-lung fibroblasts and CHO cells by radon and gamma rays of Brooks and colleagues (1994), a quality factor of 10 seems appropriate for cells at depth in the bronchial epithelium. That is half the currently recommended radiation weighting factor (ICRP 1991). It would result in partial reconciliation of dosimetrically and epidemiologically based radon risk estimates, although it would probably be misleading to over-interpret the resulting level of agreement, because many assumptions are involved in both approaches.
The mechanistic considerations discussed above must be incorporated into the design of epidemiologic analyses to estimate radon risks. We summarize the main issues of relevance to the estimation of risks associated with radon progeny.
A biologically-based risk model is a formalism that potentially provides realistic quantification of all the relevant steps from energy deposition to the appearance of cancer. If such a model were available, epidemiologic data could be fitted to it, and the resulting parameter estimates could be used to quantify the different mechanistic steps in radiation carcinogenesis. Low-dose extrapolation could then be conducted with more confidence than for a situation in which data are fitted with a purely empirical formalism. However, epidemiologic data usually include only incidence and mortality. Biologic models need cell proliferation rates and other factors, and such information is not usually available. Such approaches to radiation risk estimation have been proposed and critically dis-