Host: The Japan Radiation Research Society, Chairman of the 52nd Annual Meeting, Toshiteru Okubo (Radiation Effects Research Foundation)
The Life Span Study (LSS) of A-bomb survivors has provided the important source of risk estimates of radiation exposure in humans, and played a valuable role in establishing radiation protection system.
Since it is known that the presence of random errors in the individual radiation dose estimates for the A-bomb survivors causes underestimation of radiation effects in dose-response analyses, the method which adjusts measurement errors in dose estimates, known as regression calibration (RC), is applied in most of recent analyses using LSS data. The main reason for applying RC method is its wide applicability, and this applicability enables a great many studies using LSS data to conduct analyses in the same way.
Though it is expected that the RC method removes almost of bias in risk estimates caused by measurement errors in the individual radiation dose, the performance of RC method in LSS settings has not evaluated enough.
Therefore, we conducted a simulation study to quantify the performance of RC method. We assumed that there were two types of measurement errors inherent in the radiation dose estimates, which were classical and Berkson errors. We generated the data in a similar setting to the LSS data, and assumed Poisson regression models, and analyzed in three ways, using no adjusting method, using RC method, and using true covariate.