Statistical Methods to Study the Epidemiology of HIV & Other Disease

描述（由申请人提供）：本项目涉及发展统计技术，用于分析与艾滋病毒疾病、子宫肌瘤和其他疾病（包括性传播感染）有关的各种数据集。关于HIV，我们将主要关注用于分析HIV预防随机试验的因果推理技术。预防艾滋病毒感染的新方法的因果效应由于同时使用传统的预防方法而变得复杂。避孕套)，以及不遵守随机干预和其他同时进行的方法。其次，我们将开发具有复杂特征的时间到事件结果建模的统计方法，包括当前状态结果、竞争风险、多个有序事件或状态、测量误差以及部分已知或不完全测量的预测因子。开发这些工具的目的是分析基于横断面样本的一夫一妻制性伙伴关系中的艾滋病毒传播风险，并将其扩展到有关感染事件和暴露信息的时间和方向信息不完整的相互关联的性伙伴关系，还将推广到解释感染原因的缺失和解释在艾滋病毒传播研究中收集的暴露数据中的测量误差。基于对环境暴露于二恶英后的现有意大利妇女队列的随访信息，将扩大对多阶段疾病过程的当前状态方法的使用，以解决暴露因素与子宫肌瘤发病之间的回归关系。这些方法可以推广到艾滋病毒感染的研究，其中可以定义和利用疾病进展的各个阶段。所有这些统计工具背后的一个共同主题是，在结果和辅助因素信息不完整的情况下，对解释性因素对事件时间结果的影响进行联合和因果建模，数据本质上是观察性的，通常是高维的。这项研究将与加州大学旧金山分校的同事以及其他艾滋病毒研究人员和流行病学家合作进行，他们将允许访问他们项目的数据进行分析。这项研究与公共卫生具有相当大的相关性，因为它将允许将现代统计思想应用于以下数据的分析和解释：(1)具有复杂干预任务的随机艾滋病毒干预试验，以及(2)数据不完整的观察性研究：性伴侣之间的艾滋病毒传播，以及使用疾病诊断信息的辅助因素对疾病发生的影响。

DESCRIPTION (provided by applicant): This project concerns the development of statistical techniques for the analysis of a variety of data sets relating to HIV disease, uterine fibroids and other diseases including STIs. With regard to HIV, we will focus primarily on causal inference techniques for the analysis of randomized trials for HIV prevention. Causal effects of new methods for prevention of HIV infection are complicated by the concurrent use of traditional prevention methods (eg. condoms) in both arms of a trial, and by non-compliance with the randomized intervention and with the other concurrent methods. Second, we will develop statistical methods for modeling time-to-event outcomes with complex features including current status outcomes, competing risks, multiple ordered events or states, measurement error, and partially known or imperfectly measured predictors. These tools will be developed for the purpose of analyzing HIV transmission risk within monogamous sexual partnerships based on cross-sectional samples and extended to apply to linked sexual partnerships with incomplete information on timing and direction of infection events and exposure information, and also generalized to account for missing cause of infection and account for measurement error in exposure data collected in studies of HIV transmission. The use of current status methods for multi- stage disease processes will be extended to address regression relationships between exposure factors and the onset of uterine fibroids based on follow-up information on an existing cohort of Italian women after their environmental exposure to dioxin. These methods can be generalized to studies of HIV infection where various stages of disease progression can be defined and exploited. A common theme underlying all these statistical tools is the joint and causal modeling of the effects of explanatory factors on time-to-event outcomes when there is incomplete information on both outcomes and cofactors, with data that is observational in nature and often high dimensional. The research will be carried out in collaboration with colleagues at the University of California, San Francisco, and other HIV researchers and epidemiologists who will allow access to data from their projects for analysis. The research has considerable relevance to public health in that it will allow the application of modern statistical ideas both to the analysis and interpretation of data (i) from randomized HIV intervention trials with complex intervention assignments and (ii) from observational studies, with incomplete data, of HIV transmission between sexual partners, and the effects of cofactors on disease initiation using disease diagnosis information.