607 RADIATION RESEARCH 158, 607–614 (2002) 0033-7587/ 02 $5.00  2002 by Radiation Research Society. All rights of reproduction in any form reserved. Multistage Models and the Incidence of Cancer in the Cohort of Atomic Bomb Survivors Wolfgang F. Heidenreich, a,1 E. Georg Luebeck, b William D. Hazelton, b Herwig G. Paretzke a and Suresh H. Moolgavkar b a GSF-National Research Center for Environment and Health, Institute for Radiation Protection, 85764 Neuherberg, Germany; and b Fred Hutchinson Cancer Research Center, Public Health Sciences Division, MP-665, 1100 Fairview Avenue North, Box 19024, Seattle, Washington 98109-1024 Heidenreich, W. F., Luebeck, E. G., Hazelton, W. D., Par- etzke, H. G. and Moolgavkar, S. H. Multistage Models and the Incidence of Cancer in the Cohort of Atomic Bomb Sur- vivors. Radiat. Res. 158, 607–614 (2002). The analyses in this paper show that a number of biolog- ically based models describe cancer incidence among the A- bomb survivors equally well. However, these different mod- els can predict very different temporal patterns of risk after irradiation. No evidence was found to support the previous claim of Pierce and Mendelsohn that excess cancer risks for the solid tumors depend only upon attained age and not on age at exposure or time since exposure. Although the A- bomb survivor cohort is the largest epidemiological data set for the study of radiation and cancer, it is not large enough to discriminate among various possible carcinogenic mech- anisms. Unfortunately for hypothesis generation, the data appear to be consistent with a number of different mech- anistic interpretations of the role of radiation in carcinogenesis.  2002 by Radiation Research Society INTRODUCTION The cohort of survivors of the atomic bombings of Hi- roshima and Nagasaki continues to provide important in- sights into quantitative risk estimates from radiation car- cinogenesis in humans (1, 2). Radiation protection dose limits are based largely on estimates of cancer risk in this cohort (3). Yet a mechanistic understanding of radiation carcinogenesis continues to be important for dose and time extrapolations and for conclusions regarding health risks from other radiation qualities. Recent papers (4–7 ) have examined the data for the A-bomb survivors within the framework of multistage modeling of carcinogenesis in an effort to provide some insights into possible mech- anisms involved in the induction of cancer after irradia- tion. Questions of obvious interest are the following: Do 1 Address for correspondence: GSF-National Research Center for En- vironment and Health, Institute for Radiation Protection, 85764 Neuher- berg, Germany; e-mail: heidenreich@gsf.de. models based on ideas of multistage carcinogenesis de- scribe the data as well as empirical statistical models? Does a specific multistage model do better than others in describing the data? The first question was partially ad- dressed by Heidenreich et al. (5), who reported that mul- tistage models could describe the data just as well as the empirical models. The main objective of this paper is to address the second question. The present effort was stimulated by a recent paper of Pierce and Mendelsohn (4), which raised some provocative issues. The departure point for their analyses is the assertion that excess cancer rates for solid tumors (excluding cancers of the breast, gender-related tumors, thyroid cancers, and some minor tumors) among the people exposed to radiation in Hiroshima and Nagasaki depend upon attained age alone and not on age at exposure or time since exposure. They ask, ‘‘Can this observation help to discriminate between various models for radiation carcinogenesis? Are there gen- eral implications regarding mathematical models for carci- nogenesis?’’ They answer both questions in the affirmative and assert that the pattern of excess risks in the data can be explained if and only if (instantaneous) radiation causes the same proportionate increase in each of the transition rates of a multistage model. To address these issues, we undertook comparative analyses of cancer incidence among the A-bomb survi- vors using a number of distinct models of multistage car- cinogenesis. We used these models to analyze the inci- dence of cancer at many different sites. Many of these models fit the RERF data as well as, or better than, the Pierce-Mendelsohn model, but they predict different pat- terns of excess risk. We conclude that the data are con- sistent with many different patterns of excess risk and with different mechanisms of radiation carcinogenesis. Although in the context of epidemiological studies, the A-bomb cohort is large, cancer incidence in this cohort does not allow discrimination among various models for radiation carcinogenesis.