yet completely known. On the other hand, the duration of apparent radiation-induced increase in incidence of leukemia, which has a relatively short modal latency, is better known. Uncertainties in Dose-Response Relationships for Radiation-Induced Cancer A general hypothesis for estimation of excess cancer risk in irradiated human populations, based on theoretical considerations, on extensive laboratory animal studies, and on limited epidemiological surveys of exposed human populations, suggests various and complex doseresponse relationships between radiation dose and observed cancer incidence (NAS-BEIR, 1980). Models, with increasing complexity, include the linear, the pure quadratic, the quadratic (with a linear term), and finally, the multicomponent quadratic form with a linear term and with an exponential modifier (Figure B-l). One of the most widely considered models for cancer-induction by radiation, based on the available information and consistent with both Knowledge and theory, takes the complex quadratic form: 1(D) = (ag + aiD + a2D2)exp(-8iD-82^), where I is the cancer incidence in the irradiated population at radiation dose D in rad, and ag, a^, a2> Bl and 82 coefficients (Figure B-l). This multicomponent dose-response curve contains (1) initial upward-curving linear and quadratic functions of dose, which represent the process of cancer-induction by radiation; and (2) a modifying exponential function of dose, which represents the competing effect of cell-killing at high doses, ag is the ordinate intercept at zero dose, and defines the natural incidence of cancer in the population, ai is the initial slope of the curve at zero dose, and defines the linear component in the low-dose range. a2 determines the upward-curving quadratic function of dose, 81 and 82 determine the slope and curvature of the downward-curving function in the high-dose range, and define the cell-killing function. In the case of epidemiological surveys, this complex general dose-response form cannot be universally applied. Therefore, the model is simplified by eliminating the parameters which have the least effect on the form of the doseresponse relationship in the dose range of low-level radiation. The present SPS Committee believes: 1. Some experimental and human data, as well as theoretical considerations, suggest that for exposure to low-LET radiation, such as x-rays and gamma rays, at low doses, the linear model probably leads to overestimates of the risk of most radiation-induced cancers in man. Generally, data from high dose exposures in man and animals are used for estimating risk coefficients for the various models. To the extent that a quadratic term plays a role in actual response, high dose data extrapolated linearly to zero will tend to overestimate a, the linear term. The use of data in this fashion can certainly be used to define the upper limits of risk.
RkJQdWJsaXNoZXIy MTU5NjU0Mg==