Competing Risks Analysis Methods for Group Sequential and Adaptive Designs

项目摘要/摘要 背景：当患者可能因多种原因导致治疗失败时，使用竞争风险终点。 通过分析这些结果，可以评估治疗失败的主要原因的直接益处。 临床试验环境。回归模型可用于临床试验，以调整剩余失衡 患者特征，提高发现治疗差异的能力。但是，更有效的临床试验 设计，如分组序贯试验和适应性设计，还没有得到广泛的研究 竞争风险端点，特别是在分析中使用协变量调整的情况下。这项建议旨在 扩大具有竞争性风险结果的临床试验的设计和分析选项集，这是有限的。 具体目标：这项研究将为临床试验开发新的竞争风险方法，包括1)一组 基于Fine-Gray回归模型的疗效序贯检验，2)一组序贯检验 基于固定时间点累积发病率的直接二项回归的治疗效果，以及3)a 一种方法，允许在适应性试验中使用格雷检验，允许根据所有中期数据进行修改。 研究计划和方法：对于具体目标1和2中的每项测试，目标是推导出 成组序贯试验检验统计量序列的渐近分布。一旦此分发版本 可以选择已知的试验的提前停止边界来满足第I类错误率和功率要求。 我们将使用经验过程理论和鞅理论来得出这些结果；由此提供的强大工具 理论非常适合于时间到事件的数据，并已被用来获得固定样本和 分组序贯试验。完成特定目标3的计划涉及扩展条件的应用 拒绝原则方法(Irle&Schafer，2012)到适应性试验，使用Gray‘s检验分析 竞争风险端点。这种方法允许在维护的同时使用完整的临时数据进行修改 I类错误率。为了证明这种方法在Gray检验中的适用性，将涉及到使用鞅和 经验过程论。将进行模拟研究以验证渐近结果并检验 所提方法的有限样本性质。与NHLBI的使命相关：该项目与 NHLBI的使命是促进研究和培训，以促进心脏预防和治疗， 肺病和血液病。建议的方法将在以下试验中得到应用： 主要感兴趣的是特定类型的心脏、肺或血液疾病。该方法将在一个 一项临床试验，旨在确定一种新的药物是否能降低移植物抗宿主病的风险 血液或骨髓移植，死亡被视为相互竞争的风险。这项工作将提供更高的效率 以及灵活的设计选项，用于此类具有竞争风险的临床试验，包括使用 对协变量进行调整的治疗效果的全部中期信息和分组序贯检验。

Project Summary / Abstract Background: Competing risks endpoints are used when patients can fail therapy from several causes. Analyzing these outcomes allows one to assess the direct benefit of treatment on a primary cause of failure in a clinical trial setting. Regression models can be used in clinical trials to adjust for residual imbalances in patient characteristics, improving the power to detect treatment differences. But, more efficient clinical trial designs, such as group sequential trials and adaptive designs, have not been extensively studied with competing risks endpoints, especially when covariate adjustment is used in the analysis. This proposal aims to expand the set of design and analysis options for clinical trials with competing risks outcomes, which is limited. Specific Aims: This study will develop new competing risks methods for clinical trials including 1) a group sequential test for treatment effect based on the Fine-Gray regression model, 2) a group sequential test for treatment effect based on direct binomial regression of cumulative incidence at a fixed time point, and 3) a method permitting use of Gray's test in an adaptive trial that allows modifications based on all interim data. Research Plan and Methods: For each of the tests in specific aims 1 and 2, the goal is deriving the asymptotic distribution of the sequence of test statistics in a group sequential trial. Once this distribution is known, early stopping boundaries of the trial can be chosen to satisfy type I error rate and power requirements. Martingale and empirical process theory will be used to derive these results; the powerful tools provided by this theory are well-suited to time to event data and have been utilized to obtain methods for both fixed sample and group sequential trials. The plan for completing specific aim 3 involves application of the extended Conditional Rejection Principle approach (Irle & Schafer, 2012) to an adaptive trial using Gray's test to analyze a competing risks endpoint. This approach allows use of the full interim data for modifications while maintaining the type I error rate. Showing applicability of this method to Gray's test will involve use of martingale and empirical process theory. Simulation studies will be conducted to verify the asymptotic results and to examine the finite sample properties of the proposed methods. Relevance to the NHLBI's mission: This project aligns with the NHLBI's mission of promoting research and training to promote the prevention and treatment of heart, lung, and blood diseases. The proposed methods will find application in trials where onset of, or death from, a specific type of heart, lung, or blood diseases is the primary interest. The methodology will be illustrated on a clinical trial designed to determine whether a new agent reduces the risk of graft versus host disease after a blood or marrow transplant, where death is treated as a competing risk. This work will provide more efficient and flexible design options for clinical trials like this with competing risks, including an adaptive design with use of full interim information and group sequential testing for a treatment effect that adjusts for covariates.