Estimating Mediation and Moderation Effects in HIV Incidence Prevention Trials

项目总结/摘要 该项目的总体目标是分析旨在降低艾滋病发病率的现有行为干预措施 为了检查机制（即，介体）预测功效，以及亚组（即， 主持人）干预是有效的。我们建议使用先进的方法来分析调解人 和版主。具体而言，我们将利用直接测试间接（介导）效应的方法 并量化，即使在没有发现显著的总X-Y效应的干预措施中（即，功效）。我们提出 分析来自四项大规模随机对照试验（RCT）的数据， 干预措施对艾滋病毒发病率的影响;两项试验显示对艾滋病毒有显著影响，两项试验显示无效 艾滋病毒的结果。我们将使用新的方法称为"条件过程建模"，同时 模拟调解人和主持人，并测试干预如何以及为谁工作，帮助确定 导致不同亚组疗效的不同过程（AIM 1）。我们将比较显著和 无效试验在其重要的调解人和主持人和模型的特点，以检查可能的 无效试验（AIM 2）中设盲疗效的情况。我们将利用因果关系的尖端技术 调解分析。具体而言，我们将进行敏感性分析，以估计调解人 主要负责对艾滋病毒产生影响，加强对介体的因果推断， 在干预措施中减少艾滋病毒传播（AIM 3）。据我们所知，这些新的统计方法 从未被用于分析HIV干预RCT中的中介和调节因子。我们亦会探讨 使用生物医学策略作为介导途径的调节者，以探索报告 生物医学工具的使用显示了干预效果的不同因果过程（AIM 4）。理解 现有行为艾滋病预防干预措施的调解人和主持人将有助于告知如何预防艾滋病， 减少或增加干预措施，使其达到最有效的要素，以及哪些小组和背景必须 目标是制定更简短、更便宜和更有效的干预措施。这些干预措施可用于 与生物医学预防战略相结合。通过分析生物医学工具（即，使用nPEP， 男性包皮环切术状态）作为潜在的调解人的中介作用，我们获得了宝贵的见解， 以及当这些特定的生物医学策略 被使用。为了实现我们的目标，我们将使用“金标准”方法来分析研究数据， 艾滋病毒预防干预措施。具体来说，我们将使用来自四个不同的双臂随机对照试验的数据， 检验HIV发病率作为主要结果的能力。这些随机对照试验测试了行为艾滋病毒的有效性， 对美国男男性行为者的预防干预，南非的男人和女人， 乌干达和乌克兰的注射毒品者。因此，目前的建议将分析四个独特的 来自全球一些受艾滋病毒影响最严重的人群的干预数据集。

Project Summary/Abstract The overall goal of this project is to analyze existing behavioral interventions designed to reduce HIV incidence in order to examine the mechanisms (i.e., mediators) that predict efficacy, and the subgroups for whom (i.e., moderators) interventions are efficacious. We propose to use advanced approaches to analyzing mediators and moderators. Specifically, we will utilize methods that allow indirect (mediated) effects to be directly tested and quantified, even in interventions that do not find a significant total X ÞY effect (i.e., efficacy). We propose to analyze data from four large-scale randomized controlled trials (RCTs) testing the efficacy of behavioral interventions on an HIV incidence outcome; two trials were showed a significant effect on HIV, two had null results on HIV. We will use the novel approach called “conditional process modeling” which simultaneously models mediators and moderators and tests both how and for whom an intervention works, helping to identify different processes leading to efficacy for different subgroups (AIM 1). We will compare the significant and the null trials in their significant mediators and moderators and model characteristics to examine possible situations of masked efficacy in the null trials (AIM 2). We will utilize cutting-edge advancements in causal mediation analysis. Specifically, we will conduct sensitivity analysis to estimate the extent to which a mediator is primarily responsible for producing effects on HIV, strengthening causal inference about mediators leading to reduced HIV transmission in the interventions (AIM 3). To our knowledge these novel statistical approaches have never been used to analyze mediators and moderators in HIV intervention RCTs. We will also explore the use of biomedical strategies as moderators of mediated pathways to explore whether participants who reported use of a biomedical tool show a different causal process in intervention efficacy (AIM 4). Understanding mediators and moderators of existing behavioral HIV prevention interventions will help to inform how to pare down or ramp up interventions to their most effective elements and which subgroups and contexts must be targeted to produce briefer, cheaper, and more impactful interventions. These interventions could be used in combination with biomedical prevention strategies.!By also analyzing biomedical tools (i.e., nPEP use and male circumcision status) as potential moderators of mediated effects, we gain valuable insight into whether and which behavioral mechanisms must be targeted in interventions when these specific biomedical strategies are used. To meet our aims we will analyze data from studies using the “gold-standard” method for evaluating HIV prevention interventions. Specifically, we will use data from four different two-arm RCTs that were powered to test HIV incidence as the primary outcome. These RCTs tested the efficacy of behavioral HIV prevention interventions among men who have sex with men in the U.S., men and women in South Africa and Uganda, and people who inject drugs in Ukraine. Thus, the current proposal will analyze four unique intervention datasets from some of the most HIV-affected populations globally.