Statistical Methods and Validation Analyses for the Integration of External Data in Clinical Trials

项目摘要/摘要 我们的研究将集中于外部数据的使用，包括先前的临床研究和现实世界 在第二阶段和第三阶段肿瘤学试验的设计和分析中使用数据集。例如，我们认为， 包括根据试验和外部试验生成的数据提前停止决策的设计 患者级别的数据。 外部数据集有可能改进未来单一数据集的最终分析和临时决策 ARM和随机临床试验。他们还可以通过以下方式加快新疗法的开发 减少需要登记参加临床研究的患者数量，从而减少他们的 持续时间。然而，使用外部信息来分析临床试验目前是零星的。 事实上，集成外部患者级别的信息来测试新的治疗方法可以增加 实验治疗评估中的偏倚风险。有效地使用外部数据 临床研究的设计和分析需要充分的统计方法，以及 验证分析，以量化风险和与标准统计相比的潜在效率收益 单臂随机试验设计方案。 我们将开发新的设计，在未来的试验中使用外部数据。我们将使用数据集的集合 前列腺癌、胶质母细胞瘤和肺癌，包括患者水平的结果和预后 变量。这些收集是在临床研究中有效利用外部数据所必需的。我们会 然后引入并应用验证方法来评估使用特定疾病的统计设计 数据收集，包括临床试验和真实世界数据。验证摘要，我们将 生产，将量化试验设计的效率和整合外部数据的风险， 例如，与未测量的混杂因素或预测变量的测量误差相关 和结果。

Project Summary/Abstract Our research will focus on the use of external data, including previous clinical studies and real-world datasets, in the design and analysis of phase II and III oncology trials. We consider, for example, designs that include early stopping decisions based on data generated from the trial and external patient-level data. External datasets have the potential to improve final analyses and interim decisions of future single- arm and randomized clinical trials. They can also accelerate the development of new treatments, by reducing the number of patients that need to be enrolled in clinical studies and therefore their duration. However the use of external information to analyze clinical trials is currently sporadic. Indeed, the integration of external patient-level information to test new treatments can increase the risk of bias in the evaluation of experimental treatments. An effective use of external data in the design and analysis of clinical studies requires both, adequate statistical methodologies, and validation analyses, to quantify risks and potential efficiency gains compared to standard statistical plans of single-arm and randomized trial designs. We will develop novel designs to use external data in future trials. We will use collections of datasets in prostate cancer, glioblastoma, and lung cancer, including patient-level outcomes and prognostic variables. These collections are necessary to effectively use external data in clinical studies. We will then introduce and apply validation methods to evaluate statistical designs using disease-specific data collections, inclusive of clinical trials and real world data. The validation summaries that we will produce, will quantify the efficiency of trial designs and the risks of the integration of external data, associated for example, to unmeasured confounders or measurement errors on prognostic variables and outcome.