Non-parametric estimation of thresholds for radiation effects in vertebrate species under chronic low-LET exposures.

Databases on effects of chronic low-LET radiation exposure were analyzed by non-parametric statistical methods, to estimate the threshold dose rates above which radiation effects can be expected in vertebrate organisms. Data were grouped under three umbrella endpoints: effects on morbidity, reproduction, and life shortening. The data sets were compiled on a simple 'yes' or 'no' basis. Each data set included dose rates at which effects were reported without further details about the size or peculiarity of the effects. In total, the data sets include 84 values for endpoint "morbidity", 77 values for reproduction, and 41 values for life shortening. The dose rates in each set were ranked from low to higher values. The threshold TDR5 for radiation effects of a given umbrella type was estimated as a dose rate below which only a small percentage (5%) of data reported statistically significant radiation effects. The statistical treatment of the data sets was performed using non-parametric order statistics, and the bootstrap method. The resulting thresholds estimated by the order statistics are for morbidity effects 8.1 x 10(-4) Gy day(-1) (2.0 x 10(-4)-1.0 x 10(-3)), reproduction effects 6.0 x 10(-4) Gy day(-1) (4.0 x 10(-4)-1.5 x 10(-3)), and life shortening 3.0 x 10(-3) Gy day(-1) (1.0 x 10(-3)-6.0 x 10(-3)), respectively. The bootstrap method gave slightly lower values: 2.1 x 10(-4) Gy day(-1) (1.4 x 10(-4)-3.2 x 10(-4)) (morbidity), 4.1 x 10(-4) Gy day(-1) (3.0 x 10(-4)-5.7 x 10(-4)) (reproduction), and 1.1 x 10(-3) Gy day(-1) (7.9 x 10(-4)-1.3 x 10(-3)) (life shortening), respectively. The generic threshold dose rate (based on all umbrella types of effects) was estimated at 1.0 x 10(-3) Gy day(-1).