Analysis of lung tumour risk in radon-exposed rats: an intercomparison of multi-step modelling.

Three carcinogenesis modelling groups have both jointly and separately applied a multi-step carcinogenesis model with clonal expansion to one data set of lung tumours in rats exposed to radon (CEA, France). This study was designed to investigate the differences in modelling approach and fitting procedures used by the three groups in detail, and to explore possible discrepancies in the results. Using the same model assumptions and a (linear) radiation dependence on the first model step only, the three groups arrived at identical best fits, proving that the mathematical formalisms and fitting procedures do not lead to different results. However, when each group was allowed to find its own preferred fit for this data set, all three found a significantly better, but different fit to the data. All solutions indicated radiation to be an initiating agent and found additional radiation action necessary. The character of this additional radiation dependence, however, could not be unambiguously pinpointed. Tumour incidence data were described equally well when radiation dependence was taken into account in clonal expansion ("promotion") or in the second mutational step ("transformation"); extension to three model stages also resulted in an adequate description. The study showed that, although the three groups used one carcinogenesis model in principle, different model assumptions and/or different methods of finding the "best fit" could result in different descriptions of experimental data. This implies that on statistical grounds, different interpretations can be given for the action that radiation had in this data set. Different data, i.e. other data sets with age-dependent tumour data and/or information from cellular radiobiology experiments, are needed to specifically pin down the radiation dependence in the multi-step carcinogenesis process.