Baker Gizelle S, Hoel David G
Department of Biometry and Epidemiology at the Medical University of South Carolina, 171 Ashley Avenue, Charleston, SC 29425, USA.
Health Phys. 2003 Dec;85(6):709-20. doi: 10.1097/00004032-200312000-00016.
Cancer incidence and mortality data from the cohort of Japanese atomic bomb survivors in Hiroshima have been adjusted for the uncertainty that exists in the dose estimates, the systematic error in the neutron dose estimates, and a dose-dependent relative biological effectiveness. Once the adjustments were incorporated in the dose estimates, the data were modeled with a threshold term to allow for the possibility of a threshold dose response. The dose response models that were fit to the data were otherwise the same models used in the original papers. The threshold term was included in the model allowing for possible threshold values ranging from 0 to 0.35 Sv. These analyses suggest that the fit of the A-bomb solid tumor and leukemia incidence data are significantly improved by the addition of a threshold term in comparison with the purely linear or linear quadratic model. The results from the mortality data suggest that the leukemia data agree more with the threshold model than the linear quadratic model although the linear quadratic model is statistically equivalent, while the solid tumor data does not suggest any improvement with a threshold.
来自广岛日本原子弹幸存者队列的癌症发病率和死亡率数据，已针对剂量估计中存在的不确定性、中子剂量估计中的系统误差以及剂量依赖性相对生物效应进行了调整。一旦将这些调整纳入剂量估计中，就使用一个阈值项对数据进行建模，以考虑阈值剂量反应的可能性。拟合这些数据的剂量反应模型与原始论文中使用的模型相同。模型中包含阈值项，允许阈值在0至0.35 Sv范围内。这些分析表明，与纯线性或线性二次模型相比，添加阈值项可显著改善原子弹实体瘤和白血病发病率数据的拟合度。死亡率数据的结果表明，白血病数据与阈值模型的一致性高于线性二次模型，尽管线性二次模型在统计学上等效，而实体瘤数据未显示出阈值带来的任何改善。
