Structural models and direct and indirect effects of alcohol use on HIV infection

描述（由申请人提供）：本提案的总体目标是使用多个先前存在的nih资助的数据源来适应、应用和完善新的分析流行病学方法，以表征酒精摄入对人类免疫缺陷病毒（HIV）感染风险的关系，并估计其影响。在艾滋病毒流行开始20年后，酒精摄入与艾滋病毒血清转化之间的关系仍然知之甚少。本提案的目的将采用三项前瞻性队列研究，这些研究是美国的国宝：艾滋病与静脉注射经验（ALIVE）队列研究，多中心艾滋病队列研究（MACS）和妇女跨机构艾滋病毒研究（WIHS）。我们将对这些复杂的纵向数据调整、应用和完善边际结构模型，并描述酒精摄入量与艾滋病毒血清转化之间的关系；我们将量化酒精摄入量和艾滋病毒感染风险的特定变化的影响（例如，完全停止，克/天减半，停止暴饮）。具体来说，我们的目标是：1 .确定（1a）基于测量的行为、人口统计学和生物医学因素的酒精摄入倾向和模式，以及（1b）探索广泛（即频率和强度）和更细致（即加上类型和大小）酒精摄入评估之间的关系；2-估计酒精摄入和模式对艾滋病毒血清转化风险的总影响（即直接和间接影响），探讨总影响是否受年龄、性别或日历年的影响，并量化由于测量误差、未测量混杂和选择偏差造成的不确定性；估计酒精摄入和模式以及非法或娱乐性处方药使用对艾滋病毒血清转化风险的共同（即修正）影响；4-探索酒精摄入和模式对艾滋病毒血清转化风险的间接影响，通过随后的(a)危险的性行为或(b)测量的免疫功能标志物介导。将个人随机分配到对公众健康重要的多种暴露（如酒精摄入）的特定水平，既不道德又不切实际。时变混杂、测量误差、迁出性选择偏差（由于随访损失）、时变风险行为/免疫功能的调解以及效应修正等问题都有效地适用于酒精摄入和HIV血清转化的观察数据，并在本提案中加以解决。拟议的研究与生物医学和公共卫生研究中真正的、长期的挑战直接相关。具体而言，研究结果将为制定针对艾滋病毒感染风险人群的有针对性的干预措施和公共卫生信息提供必要的证据。总之，本提案描述了一个独特的机会和一个重要的、创新的和具有成本效益的倡议，在一个重要的科学研究领域的背景下全面检查、应用和完善最先进的分析方法。公共卫生相关性：拟议的研究将以两种方式与生物医学和公共卫生研究中的实际、长期挑战直接相关。首先，它将为理解酒精摄入与人类免疫缺陷病毒感染之间的关系提供必要的清晰度。第二，为复杂观测数据分析的定量方法提供创新和详细的工作实例。

DESCRIPTION (provided by applicant): The overall goal of this proposal is to use multiple preexisting NIH-funded data sources to adapt, apply and refine novel analytic epidemiologic methods to characterize the relation, and estimate the impact, of alcohol intake on risk of infection with Human Immunodeficiency Virus (HIV). The relation between alcohol intake and HIV seroconversion remains poorly understood >2 decades after the onset of the HIV epidemic. The aims of the present proposal will be undertaken using three prospective cohort studies, which are US national treasures: the AIDS Link to Intravenous Experience (ALIVE) cohort study, the Multicenter AIDS Cohort Study (MACS) and the Women's Interagency HIV Study (WIHS). We will adapt, apply and refine marginal structural models to these complex longitudinal data and characterize the relation between alcohol intake and HIV seroconversion; we will quantify the impact of specified changes in alcohol intake and risk of HIV infection (e.g., complete cessation, halving of grams/day, cessation of binging). Specifically, we aim to: 1-determine the (1a) propensity of alcohol intake and patterns based on measured behavioral, demographic and biomedical factors, and (1b) explore the relation between broad (i.e., frequency and intensity) and more nuanced (i.e., plus type and size) alcohol intake assessments; 2- estimate the total effect (i.e., direct and indirect) of alcohol intake and patterns on risk of HIV seroconversion, explore whether the total effect is modified by age, sex or calendar year, and quantify the uncertainty due to measurement error, unmeasured confounding, and selection bias; 3-estimate the joint (i.e., modifying) effect of alcohol intake and patterns and illicit or recreational prescription drug use on risk of HIV seroconversion; and 4- explore the indirect effects of alcohol intake and patterns on risk of HIV seroconversion, mediated through subsequent (a) risky sexual behavior or (b) measured markers of immune function. Randomization of individuals to specified levels of many exposures important to the public's health (e.g., alcohol intake) is both unethical and impractical. Issues of time-varying confounding, measurement error, emigrative selection bias (due to follow up loss), mediation by time-varying risk behaviors/immune function, and effect modification all apply in force to observational data on alcohol intake and HIV seroconversion and are addressed in this proposal. The proposed research is directly relevant to real, long-standing challenges in biomedical and public health research. Specifically, research findings will provide necessary evidence for the development of targeted interventions and public health messages in persons at-risk for HIV. In summary, the present proposal describes a unique opportunity and a significant, innovative and cost-effective initiative to comprehensively examine, apply and refine start-of-the-art analytic methods in the context of an important scientific field of inquiry. PUBLIC HEALTH RELEVANCE: The proposed research will be directly relevant to real, long-standing challenges in biomedical and public health research in two ways. First, it will provide required clarity in understanding the relation between alcohol intake and Human Immunodeficiency Virus infection. Second, it will provide innovations and detailed worked examples in quantitative methods for complex observational data analysis.