Krewski Daniel, Burnett Richard T, Goldberg Mark S, Hoover B Kristin, Siemiatycki Jack, Jerrett Michael, Abrahamowicz Michal, White Warren H
McLaughlin Centre for Population Health Risk Assessment, Ottawa, Ontario, Canada.
J Toxicol Environ Health A. 2003;66(16-19):1507-51. doi: 10.1080/15287390306424.
This article provides an overview of the Reanalysis Study of the Harvard Six Cities and the American Cancer Society (ACS) studies of particulate air pollution and mortality. The previous findings of the studies have been subject to debate. In response, a reanalysis team, comprised of Canadian and American researchers, was invited to participate in an independent reanalysis project to address the concerns. Phase I of the reanalysis involved the design of data audits to determine whether each study conformed to the consistency and accuracy of their data. Phase II of the reanalysis involved conducting a series of comprehensive analyses using alternative statistical methods. Alternative models were also used to identify covariates that may confound or modify the association of particulate air pollution as well as identify sensitive population subgroups. The audit demonstrated that the data in the original analyses were of high quality, as were the risk estimates reported by the original investigators. The sensitivity analysis illustrated that the mortality risk estimates reported in both studies were found to be robust against alternative Cox models. Detailed investigation of the covariate effects found a significant modifying effect of education and a relative risk of mortality associated with fine particles and declining education levels. The study team applied spatial analytic methods to the ACS data, resulting in various levels of spatial autocorrelations supporting the reported association for fine particles mortality of the original investigators as well as demonstrating a significant association between sulfur dioxide and mortality. Collectively, our reanalysis suggest that mortality may be attributable to more than one component of the complex mixture of ambient air pollutants for U.S. urban areas.
本文概述了哈佛六城市再分析研究以及美国癌症协会（ACS）关于颗粒物空气污染与死亡率的研究。这些研究先前的发现一直存在争议。作为回应，一个由加拿大和美国研究人员组成的再分析团队受邀参与一个独立的再分析项目，以解决这些担忧。再分析的第一阶段涉及设计数据审核，以确定每项研究是否符合其数据的一致性和准确性。再分析的第二阶段涉及使用替代统计方法进行一系列综合分析。还使用替代模型来识别可能混淆或改变颗粒物空气污染关联的协变量，以及识别敏感人群亚组。审核表明，原始分析中的数据质量很高，原始研究人员报告的风险估计也是如此。敏感性分析表明，两项研究中报告的死亡率风险估计对替代考克斯模型具有稳健性。对协变量效应的详细调查发现教育有显著的修正效应，以及与细颗粒物和教育水平下降相关的死亡率相对风险。研究团队将空间分析方法应用于ACS数据，得出了不同程度的空间自相关性，支持了原始研究人员报告的细颗粒物死亡率关联，同时也表明二氧化硫与死亡率之间存在显著关联。总体而言，我们的再分析表明，在美国城市地区，死亡率可能归因于环境空气污染物复杂混合物的不止一个成分。
