This is a review of historical trends, theoretical concepts and experimental data pertaining to the question of the shape of the dose-response curve for carcinogenesis in human beings exposed to low LETradiation. The focus of the study is on the question of whether linear extrapolation data obtained at highdosesand dose ratesisareasonable orunreasonable method of evaluating risk of concer development following irradiation at low doses and/or low dose rates.Most, but not all, of the animal data indicate that linear extrapolation from high doses overestimates the risk at low doses or dose rates. However, a review of the mechanisms and possible dose responsecurves foreach,suggests that thisgeneral principle may not be applicable to human radiation carcinogenesis. An analysis of dose-response curves for two-hit chromosome aberrations in mammalian cells shows that there is a linear component which dominates the response up to approx 100 rads. Contrary to popular belief there are quite good data on cancer induction in humans by low doses and/or low dose rates of low LET radiation. Risk estimates at these low doses were found not to be less than comparable risk estimates made from highdoses. If anything, in fact, the data suggest the opposite.It is concluded that linear extrapolation of human data from high doses (of up to around 100 rads) to low doses of low LET radiation may not overestimate the risk at low doses. Linear extrapolation of high LET radiation from high to low doses probably underestimates the risk at low doses in some cases.
HISTORICAL PERSPECTIVETHE LINEAR. non-threshold model for estimating the carcinogenic risk of a given dose of ionizing radiation was developed and used subsequent to 1950. Prior to this time it had been assumed (at least for standard-setting purposes) that cancer, like other somatic effects such as erythema, epilation and lethality, was produced by radiation according to nonstochastic kinetics. That is, a certain level of damage had to be produced to initiate the event, and so the dose-response curve would be sigmoidal with a threshold (Fig. 1). It was the accumulating data on the induction of mutations in the fruit fly Drosophila melanognster which were well-fitted by a linear non-threshold dose response curve, plus the need to make quantitative estimates of the *