Obviously some specific issues would have to be considered in such models, including the dose and dose rate effectiveness factor (DDREF) and the shape of the dose-response curve, the temporal distribution of risk after exposure, and the interaction of radiation with other risk factors and with other possible modifying factors, such as sex, age at exposure, attained age, and population differences.

Approaches to the modeling process could include:

  • Fitting of purely empirical models to original data from studies or combined studies.

  • Fitting of purely empirical models with meta-analysis; this is relatively underdeveloped and might be particularly useful when there are a number of studies of a particular outcome such as esophageal cancer.

  • Fitting semiempirical biologically based models to epidemiologic data to improve understanding of the biologic basis of some of the empirical effects observed.

  • Fitting (and testing) of simple models now being used in radiation protection, such as linear nonthreshold models in which the estimated relative risk at 1 sievert might depend upon age at exposure but remains invarient over time after exposure (with a minimal latent period) or an otherwise similar quadratic (linear-quadratic) model with an appropriate DDREF and particular attention given to the principal contending alternatives. Such alternatives include hormesis, threshold models, the Kellerer-Barclay model, and supralinear models.

Committee members will be selected who will be able to access original data from completed or ongoing studies, or who will be able to directly contact the original investigators.


The NIH radioepidemiologic tables, mandated by Congress, were developed to meet a perceived need for an objective way to present and evaluate compensation claims for adverse health outcomes, such as cancer, that might be related to radiation exposure. The concept is simple: given a documented history of exposure to radiation d1, . . ., dk at ages a1, . . ., ak and a cancer diagnosis at age A, compute the ERR of a cancer at that age. The ''probability of causation" (NIH 1985), or "assigned share" (NRC 1984), computed as ERR/(1 + ERR), is an informed quantitative estimate of the proportion of similar cancers at that age, in a large population of similar people with similar exposure histories that, would not have occurred in the absence of exposure, that is, the proportion of such cancers attributable to radiation. The ERR might depend on exposure history and age at diagnosis, but also on sex, time from each exposure until diagnosis, history of exposure to other carcinogens (such as tobacco), and other risk modifiers (such as reproductive history). Thus, all relevant factors known to influence radiation-related risk can be incorporated, as can various sources of uncertainty.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement