Anderson Jeri L, Apostoaei A Iulian
Division of Surveillance, Hazard Evaluations and Field Studies (DSHEFS), National Institute for Occupational Safety and Health (NIOSH), Cincinnati, Ohio, USA.
Oak Ridge Center for Risk Analysis, Oak Ridge, Tennessee, USA.
J Expo Sci Environ Epidemiol. 2017 Jan;27(1):1-6. doi: 10.1038/jes.2015.36. Epub 2015 May 13.
In retrospective epidemiological studies of large cohorts of workers exposed to radioactive materials, it is often necessary to analyze large numbers of bioassay data sets containing censored values, or values recorded as less than a detection limit. Censored bioassay data create problems for all bioassay analysis methods, including analytical techniques based on least-squares regression to estimate intakes. A method is presented here that uses a simple empirically-derived equation for imputing replacement values for urine uranium concentration results reported as zero or less than a detection limit, that produces minimal bias in intakes estimated using least-square regression methods with the assumption of lognormally distributed measurement errors.
在对大量接触放射性物质的工人队列进行回顾性流行病学研究时，常常需要分析大量包含截尾值（即记录为小于检测限的值）的生物测定数据集。截尾生物测定数据给所有生物测定分析方法都带来了问题，包括基于最小二乘回归来估算摄入量的分析技术。本文提出了一种方法，该方法使用一个简单的经验推导方程来为报告为零或小于检测限的尿铀浓度结果估算替代值，在假设测量误差呈对数正态分布的情况下，使用最小二乘回归方法估算摄入量时，该方法产生的偏差最小。
