The epidemiology of radiation carcinogenesis is likely to be dominated during the next 25 y by the gradual maturation of major studies of populations exposed to substantial levels of radiation dose. During this period, for example, most of the new information on cancer risk among survivors of the 1945 atomic bombings of Hiroshima and Nagasaki will pertain to persons exposed at ages under 30. From the A-bomb survivor data and from a number of medically and occupationally irradiated populations now under study, it will be possible to construct a fairly complete picture of the temporal distribution of excess cancer risk over the remaining life span following exposure, a major source of uncertainty and controversy in current risk estimates. Provided that the necessary efforts are made, it should be possible to greatly increase our understanding of which organs and tissues are vulnerable to radiation carcinogenesis, variation by age at exposure and sex, and (what is most difficult at the present time) the influence of other risk factors and possible modifying factors on the magnitude of risk associated with radiation exposure. For some sites it may be possible, as it is now for breast cancer, to rule out certain broad classes of simple dose-response models. While the data gathered during the next 25 y will be nearly the last from some very important series, their information content should be extremely high. Thus, there is every reason to improve the resources, e.g., by improving dosimetry, enlarging samples, tapping new sources of case ascertainment, and gathering information on factors other than radiation dose. Modeling in epidemiological investigations depends heavily on work in experimental radiobiology. Thus far, the exchange has been largely in one direction, since epidemiological study design is mostly a matter of extracting information from “experiments” that have already taken place. With more epidemiological information, it seems reasonable to hope that detailed hypotheses, with direct relevance to areas of epidemiological uncertainty, may be generated suitable for experimental investigation. It seems likely that findings from large studies of populations exposed to relatively low levels of radiation, for which the signal-to-noise ratio is low, and for which random error and subtle sources of bias may have more than usual influence on reported results, will continue to generate controversy. Here, there seems no alternative to patient and persistent efforts to improve the general level of understanding of the principles of scientific inference, so that the seemingly more direct interferences from low-dose studies can be viewed in their proper perspective.
©1988Health Physics Society