Statistical Methods to Account for Exposure Uncertainty in Environmental Epidemiology

 描述(由申请人提供)：暴露测量误差可能是几乎所有环境健康研究中的偏差来源，通常导致低估相对风险和丧失检测影响的统计能力。在这项拟议的研究中，我们将采取与NIEHS战略优先事项一致的生命过程方法，重点关注环境健康几个关键领域的方法学需求，包括空气污染成分和邻近环境因素对心血管疾病及其前驱和后果的影响，包括全因死亡率、肥胖症、2型糖尿病和亚临床心血管生物标志物。我们已经组建了一支由领先的理论和应用统计学家、环境流行病学家和环境暴露评估专家组成的强大的多学科团队，与NIEHS的另一个战略目标保持一致，我们将在具有紧迫环境卫生政策重要性的新领域做出重大贡献，主要重点是开发方法，准确量化复杂的单一和多个、跨空间和跨时间的同时暴露影响的影响，响应NIEHS的另一个战略优先事项，减少(如果不是消除)由于存在重大暴露测量误差而存在的偏差和效率损失。在这项工作中，将仔细注意消除空间和时间混杂造成的偏差，以及调整室内空气污染源的混杂。目前可获得的关于空气污染成分和邻近环境特征的验证数据将被汇编起来，并用于开发与手头数据相适应的个人暴露与测量的环境暴露之间的测量误差模型。混合纵向和COX生存数据回归模型将构成分析框架的基础。实施这些方法的用户友好的软件将在网上张贴，促进新方法在广泛的环境健康问题上的广泛应用。

 DESCRIPTION (provided by applicant): Exposure measurement error is a likely source of bias in nearly all environmental health studies, typically leading to an under-estimation of relative risks and a loss of statistical power to detect effects. In this proposed research, we wil take a life course approach, as consistent with NIEHS strategic priorities, focusing on methodological needs in several critical areas of environmental health, including the effects of constituents of air pollution and of aspects of the neighborhood environment on cardiovascular disease and its precursors and consequences, including all-cause mortality, obesity, type 2 diabetes and subclinical cardiovascular biomarkers. Having assembled a strong multi-disciplinary team of leading theoretical and applied statisticians, environmental epidemiologists and environmental exposure assessment experts, consistent with another NIEHS strategic objective, we will make significant contributions to novel areas of pressing environmental health policy importance, with a major focus on developing methods to accurately quantify the effects of complex single and multiple, simultaneous exposure effects across space and time, responding to another NIEHS strategic priority, reducing if not eliminating the bias and loss of efficiency otherwise present due to the presence of substantial exposure measurement error. In this work, careful attention will be paid to removing bias due to spatial and temporal confounding as well as to adjusting for confounding by indoor sources of air pollution. Currently available validation data on air pollution constituents and on features of the neighborhood environment will be assembled and used to develop measurement error models relating personal exposure to measured ambient exposure as suitable for the data at hand. Mixed longitudinal and Cox survival data regression models will underlie the analytic framework. User-friendly software implementing the methods will be posted on the web, facilitating wide-scale application of the new methods to a broad range of environmental health problems.