Tests of the linearity assumption in the dose-effect relationship for radiation-induced cancer

The validity of the BEIR linear extrapolation to low doses of the dose-effect relationship for radiation induced cancer is tested by use of natural radiation making use of selectivity on type of cancer, smoking habits, sex, age group, geographic area and/or time period. For lung cancer, a linear interpolation between zero dose-zero effect and the data from radon-induced cancers in miners implies that the majority of all lung cancers among non-smokers are due to radon; since lung cancers in miners are mostly small-cell undifferentiated (SCU), a rather rare type in general, linearity over predicts the frequency of SCU lung cancers among non smokers by a factor of 10, and among non-smoking females age 25-44 by a factor of 24. Similarly, linearity predicts that the majority of all lung cancers early in this century were due to radon even after due consideration is given to cases missed by poor diagnostic efficiency (this matter is considered in some detail). For the 30-40 age range, linearity over predicts the total lung cancer rate at that time by a factor of 3-6; for SCU lung cancer, the over-prediction is by at least a factor of 10. Other causes of lung cancer are considered which further enhance the degree to which the linearity assumption over-estimates the effects of low level radiation. A similar analysis is applied to leukemia induced by natural radiation. It is concluded that the upper limit for this is not higher than estimates from the linearity hypothesis. The use of the selectivities is justified by the fact that the total incidence of lung cancer or leukemia is an upper limit on the rate at which it is caused by natural radiation effects; in determining upper limits it is justifiable to select situations which minimize them.