Statistical Methods for Advancing HIV Prevention

项目概要/摘要 基于抗逆转录病毒的HIV预防在预防HIV感染方面非常有效，但仍有许多 执行方面的挑战。以人群为基础的有效性研究旨在评估预防性治疗对现实世界的影响。 干预措施。然而，我们目前衡量预防性措施有效性的能力存在重大限制。 在人口一级的干预措施，因为它需要资源广泛的纵向测试，在一个封闭的 队列。HIV近因检测（提供HIV感染时间信息的检测）提供了资源- 有效估计发病率。然而，由于缺乏精确度，这种测定的实用性目前受到限制。在 此外，设计有效性试验以评估新的艾滋病毒预防干预措施越来越具有挑战性， 存在有效的预防剂。为了填补这些空白，我们将改进衡量艾滋病毒的统计方法 发病率，并开发一种新的试验设计，以评估艾滋病毒预防性干预措施的有效性。具体来说，我们将 扩展现有方法，使用近因分析数据估计HIV发病率，以适应协变量效应 检测特性、HIV发病率的时间趋势，以及更精确地估计HIV发病率。 我们还将开发一种新的艾滋病预防有效性试验设计，称为“增强的活性对照试验”。 设计“，将利用额外的信息来推断艾滋病毒的发病率没有干预，即”反事实“ 安慰剂组的HIV发病率。为了扩展和发展这些方法，我们将定义一个统计框架;定义 估计和推断给定数据的参数的方法;推导和比较分析 推理方法的性质;评估模拟研究中的性能;并将方法应用于真实的 数据来产生新的科学见解。这些新方法可直接应用于评估 基于人群的有效性研究和随机对照有效性试验中的艾滋病毒预防干预措施， 并将适用于其他传染病的研究。 1

Project Summary/Abstract Antiretroviral-based HIV prophylaxis is highly effective at preventing acquisition of HIV, yet there are many implementation challenges. Population-based effectiveness studies seek to evaluate real-world impact of preventive interventions. There is, however, a major limitation in our current ability to measure effectiveness of preventive interventions at the population-level due to its requirement of resource-extensive longitudinal testing in a closed cohort. HIV recency assays (assays that provide information on the timing of HIV acquisition) offer resource- efficient estimates of incidence. However, the utility of such assays is currently limited due to lack of precision. In addition, designing efficacy trials to evaluate new HIV preventive interventions is increasingly challenging when effective prevention agents exist. To fill these gaps, we will advance statistical methodology to measure HIV incidence and develop a new trial design to assess efficacy of an HIV preventive intervention. Specifically, we will extend existing methods for estimating HIV incidence using recency assay data to accommodate covariate effects on assay properties, temporal trends in HIV incidence, and to estimate HIV incidence with increased precision. We will also develop a new class of HIV prevention efficacy trial design termed the ‘augmented active-controlled design’ which will leverage additional information to infer HIV incidence absent intervention, i.e. ‘counterfactual placebo’ HIV incidence. To extend and develop these methods, we will define a statistical framework; define approaches to estimating and drawing inference about parameters given the data; derive and compare analytic properties of the inferential methods; evaluate performance in simulation studies; and apply the methods to real data to generate new scientific insights. These novel methods have direct application to evaluating the impact of HIV preventive interventions in population-based effectiveness studies and randomized controlled efficacy trials, and will be applicable to the study of other infectious diseases. 1