Accounting for Berkson and Classical Measurement Error in Radon Exposure Using a Bayesian Structural Approach in the Analysis of Lung Cancer Mortality in the French Cohort of Uranium Miners
Accounting for Berkson and Classical Measurement Error in Radon Exposure Using a Bayesian Structural Approach in the Analysis of Lung Cancer Mortality in the French Cohort of Uranium Miners
复制标题
DOI：
10.1667/rr14467.1
复制
发表时间：
2017-02-01
影响因子：
3.4
通讯作者：
Ancelet, Sophie
中科院分区：
文献类型：
--
作者：
Hoffmann, Sabine;Rage, Estelle;Ancelet, Sophie
许多关于地下矿工的职业队列研究表明，氡暴露与肺癌死亡风险增加有关。然而，尽管曝光测量误差对统计推断的有害后果，这些分析传统上并不考虑曝光的不确定性。这可能是由于不完善的氡暴露替代措施造成的测量误差具有挑战性。事实上，我们通常在一个随时间变化的曝光变量中面临曝光的不确定性，其中误差的类型和幅度可能取决于曝光的时间段。为了应对对可能具有Berkson或经典性质的乘性和异方差测量误差进行核算的挑战，根据暴露年份的不同，我们选择了贝叶斯结构方法，这可以说是解释暴露评估中不确定性的最灵活的方法。我们在法国铀矿工人队列中评估了职业氡暴露与肺癌死亡率之间的关系，发现不相关的乘性测量误差的影响微乎其微。然而，我们的调查结果表明，在这一群体以及许多其他地下矿工群体采矿的最初几年进行的接触评估的回溯性可能会导致暴露-风险关系的减弱。需要更多的研究来解决肺部剂量计算中的进一步不确定性，因为这一步骤可能会引入共同不确定性的重要来源。(C)2017年，由辐射研究学会提供
Many occupational cohort studies on underground miners have demonstrated that radon exposure is associated with an increased risk of lung cancer mortality. However, despite the deleterious consequences of exposure measurement error on statistical inference, these analyses traditionally do not account for exposure uncertainty. This might be due to the challenging nature of measurement error resulting from imperfect surrogate measures of radon exposure. Indeed, we are typically faced with exposure uncertainty in a time-varying exposure variable where both the type and the magnitude of error may depend on period of exposure. To address the challenge of accounting for multiplicative and heteroscedastic measurement error that may be of Berkson or classical nature, depending on the year of exposure, we opted for a Bayesian structural approach, which is arguably the most flexible method to account for uncertainty in exposure assessment. We assessed the association between occupational radon exposure and lung cancer mortality in the French cohort of uranium miners and found the impact of uncorrelated multiplicative measurement error to be of marginal importance. However, our findings indicate that the retrospective nature of exposure assessment that occurred in the earliest years of mining of this cohort as well as many other cohorts of underground miners might lead to an attenuation of the exposure-risk relationship. More research is needed to address further uncertainties in the calculation of lung dose, since this step will likely introduce important sources of shared uncertainty. (C) 2017 by Radiation Research Society