ANS National
 Meetings/
 Sessions

November 1995

Joseph L. Alvarez

(lT Corp, Englewood)

And

Fritz A. Seiler

(IT Corp, Albuquerque)

5. Why We Need New Approaches to Low-Dose Risk Modeling

                “The linear no-threshold model for radiation effects was introduced by the International Commission on Radiological Protection as a conservative model for the design of radiation protection programs.  'The model has persisted not only as the basis for such programs but has come to be treated as a dogma and is often confused with Scientific fact. The pervasive use of the linear model has extended calculations of radiation risks to dose levels that cannot be measured and have risks that are much smaller than the uncertainty of the calculation. This extends the model beyond its range of applicability and requires consideration of a minimum significant risk.

                “Examinationof the model for a minimum significant risk required examination of the model and data uncertainties. In this examination a number of serious problems with the linear no-threshold model of radiation carcinogenesis were demonstrated, many of them invalidating the hypothesis. It was shown that the relative risk formalism did not approach one as the dose approaches zero. When mortality ratios were used instead, the data in the region below 0.3 Sv were systematically below the predictions of the BEIR V model. It was also shown that the data above 0.3 Sv were of little use in formulating a model at low doses. In addition, these data are valid only for doses accumulated at high dose rates, and there is no scientific justification for using the model in low-dose, low-dose-rate extrapolations for purposes of radiation protection. Figure I illustrates the state of our present knowledge from the data base of the BEIR V model and the regions of dose and dose rate that are dependent on the model for interpretation. The dark shaded portion of the figure is the region of the Japanese survivors, while the light shaded portion is the low-dose and low-dose-rate region. Data exist in the intermediate region to help bridge the considerable gap from the region of the model to the low-dose and low-dose-rate region. Above all, the high-dose to low-dose model ignores the copious epidemiological database that is available at low doses and dose rates, that is, exactly the region for which radiation protection models are needed. These epidemiological data also are at odds with the linear no-threshold hypothesis.

                “Despitethese obvious errors in the model extrapolation, further examination of model fits to the Japanese survivor data were attempted.  Several such models were fit to the data including an unconstrained linear, linear-square root, and Weibull, all of which fit the data better than the relative-risk, linear no-threshold model. These fits were used to demonstrate that the BEIR V model systematically overestimates the risk at low doses in the Japanese survivor data set. This systematic lack of fit hardly invites extension to low rates.

                “Itis recommended that an unbiased reanalysis of the data be undertaken and the results used to construct a new model, based on all pertinent data. This model could then form the basis for managing radiation risks in the appropriate dose and dose rate.”

1."Recommendations of the International Commission on Radiological Protection," ICRP Publication 26, Pergamon Press, Oxford (1977).

2.Health Effects of Exposure to Low Levels Ionizing Radiation: BEIR V, National Academy Press, Washington, DC (1990).

3."Recommendations of the International Commission on Radiological Protection, "ICRP Publication 60, Pergamon Press, Oxford (1991).

4. E A.SEILER, J. L. ALVAREZ, "The Definition of a Minimum Significant Risk," Technology, 331A, 83 (1994).