Methods to improve the detection of treatment effects in ARDS clinical trials

 描述(由申请人提供)：每年约有140,000名美国人患上急性呼吸窘迫综合征(ARDS)，其中40%的患者死亡。然而，在所有旨在检测死亡率降低的ARDS随机临床试验(RCT)中，只有5%能够确定新的、成功的干预措施。除了提高存活率，2009年NHLBI研讨会确定了未来ARDS随机对照试验的几个优先事项，包括制定招收更少患者的战略，以及关注ARDS幸存者以患者为中心的长期终点。虽然重症监护界试图确定干预措施对重症监护病房(ICU)后终点的影响，如生活质量(QOL)和身体和认知功能，但在评估ARDS患者的非致命终点方面存在几个方法学挑战。这些问题包括高患者死亡率(即，对死亡进行审查)导致的数据缺失、ARDS患者子集之间的不同治疗效果，以及缺乏以患者为中心的长期结果的有效的早期替代终点。这项研究和培训计划的总体目标是执行三项相互关联的研究，作为临床流行病学博士论文的主题，这三项研究将对改善ARDS患者随机对照试验结果的进行、分析和应用具有直接意义。首先，在目标1中，我们将提出一个创新的统计框架，用于分析短期和长期非致命后果，以解决因死亡而导致的数据缺失造成的偏差。具体地说，我们将利用具有纵向和生存组件的联合建模框架，从而提高统计能力(潜在地降低 随机对照试验中所需的样本量)，同时比随机对照试验中常用的方法提供更大的临床和统计推断。接下来，在目标2中，我们将基于患者级别的数据生成多变量风险评分，该评分可用于指导不同类型ARDS患者的更多个性化治疗建议，并将其集成到我们的联合建模框架中。最后，在目标3中，我们将寻求确定ICU通风暴露的阈值效应(例如，需要机械通风的时间、无呼吸机的天数)，以预测长期 死亡率与ICU后生活质量。严格的课程，包括在危重病护理、统计学和高级流行病学方面的授课和经验式学习，将完善申请人的培训，使他在完成博士学位后成为一名独立和合作的教研员和危重病护理试验人员。

 DESCRIPTION (provided by applicant): Each year, approximately 140,000 Americans develop acute respiratory distress syndrome (ARDS), and 40% of these patients die. Yet, only 5% of all ARDS randomized clinical trials (RCTs) designed to detect reductions in mortality have been able to identify new, successful interventions. In addition to improving survival, a 2009 NHLBI workshop identified several priorities for future ARDS RCTs, including the development of strategies to enroll fewer patients and a focus on longer-term patient-centered endpoints for ARDS survivors. Although the critical care community has sought to identify interventions' effects on post intensive care unit (ICU) endpoints such as quality of life (QOL) and physical and cognitive functioning, there are several methodological challenges in assessing non-mortal endpoints among ARDS patients. These include missing data resulting from high patient mortality rates (i.e., censoring from death), heterogeneous treatment effects among subsets of ARDS patients, and the lack of validated, early surrogate endpoints for long-term patient-centered outcomes. The overall goal of this research and training plan is to execute three interrelated studies as the subject of a doctoral dissertation in clinical epidemiology that will have immediate relevance in improving the conduct, analysis and application of findings from RCTs for ARDS patients. First, in Aim 1, we will advance an innovative statistical framework for the analysis of short- and long-term non-mortal outcomes that addresses the biases resulting from death-induced missing data. Specifically, we will utilize a joint modeling framework with longitudinal and survival components, resulting in gains in statistical power (potentially reducing the needed sample sizes in RCTs), while offering greater clinical and statistical inference than current methods commonly used in RCTs. Next, in Aim 2, we will generate a multivariable risk score based on patient-level data that can be used to guide more individualized treatment recommendations among heterogeneous ARDS patients, and integrate it into our joint modeling framework. Finally, in Aim 3, we will seek to identify threshold effects of ICU ventilation exposure (e.g., time requiring mechanical ventilation, ventilator-free days) that predict long-term mortality and post-ICU QOL. A rigorous curriculum including didactic and experiential learning in critical care, statistics, and advanced epidemiology will round out the applicant's training, preparing him to be an independent and collaborative faculty investigator and critical care trialis at the completion of his PhD.