906 J. A. Reissland et al. these datait is not possible to predict the eventual shape of the time distribution of the cancer rate. To estimate the absolute risk of radiation-indueced cancer death, Goss doubled the number of excess cancers which had occurred up to 1970, This leads to a risk coefficient of 70 per 10° man rads and hence to a total of 100 per 106 manrads for all cancers (including leukaemia). These figures apply to gamnia radiation. Risk coefficients obtained from the study of other irradiated groups are not easily related to the conditions of whole body exposure experienced by the occupationally exposed workers of interest to us. for example, radiotherapy patients such as ankylosing spondylitics (Court Brown and Doll 1965) and those treated for metropathia haemorrhagica (Smith and Doll 1976) receive high and localized exposures which are very dissimilar to those of our group ancl hence may have a quite different excess canccr pattern. Also, those exposed are a special group who may exhibit abnormal medical response to the exposure whereas we are concerned with a predominantly healthy group. Studies of the radiologists in the USA (Matanoski, Seltser, Sartwell, Diamond and Elliott 1975) are probably the most comparably exposed group to the classified workers in our study, bub unfortunately their doses are not recorded andso risk coefficients cannot be deduced, neitheris it possible to extract the time pattern of their excess cancers. Tor the purposes of the present paper it is proposed to use a figure of 100 cancer deaths per 10° manrads (1074 per rad), based on the Japanese survivor data corrected for gamma-ray exposure only. With the same justification we use a risk coefficient of 30 per 10° manrads (3x 10-5 per vad) for excess loukaemias. These figures may be regarded as conservative cstimates when applied to the low doses received at low dose rates by radiation workers. Biological repair mechanisms will act to reduce radiation damage to tissue Lo belowthat expected from high cose rate observations. This may he taken into account by a protraction factor but our knowledge of low dose rate effects is inadequate to establish a value for this factor so it will be assumed that the risk is lincarly related to the dose for the rauge of doses accumulated by radiation workers. So for example, lrad accumulated by each of LO® persons will lead to the same numberof cancer deaths as 100 rad to cach of 10" persons. Phe time variation of the number of execss cancer mortalitics following exposure is not clear from the currently available evidence auc so we have made an arbitrary choice for our calculations. We have assumed that the risk occurs over a limited period rather than remaining ata high valuc tor all times after exposure. The Japanese data indicate an increased leukaemia risk lasting about 30 years and a simple assumption is that the risk remains constant over the period 5-30 years after exposure. Forall cancers the risk appears to exist for a longer time andib is assumed constant from 5 to 50 years. In orderto test the sensitivity of our results to the assumed form ofthe risk with time, we have cousidered three forms ofrisk-in-time following exposure: A, a rectangular distribution, ‘B, a Gaussian distribution, and CF C, asharply peaked disteibution—all the risk during the 11th yeav.