Innovative statistical methodologies to subgroup analysis in clinical trials

FDA PA 19-306 Bierer，Barbara E.，M.D. 临床试验亚组分析的创新统计方法 项目摘要 布里根妇女医院和哈佛大学多区域临床试验(MRCT)中心 中心)是一个研究和政策中心，旨在解决行为、监督、道德和监管问题 临床试验的环境，重点是多国试验。为了完成这项工作，我们的职能是 独立召集人，吸引来自工业界、CRO、学术界、患者和 患者权益倡导团体、非营利组织和全球监管机构，以解决问题 审判的严密性和完整性。在这项建议中，我们建议召开一次公开会议 统计学家、临床试验专家、监管者和患者/患者倡导者讨论创新的统计学 临床试验中亚组分析的方法学。这次会议与FDA的 努力促进将不同背景和特征的个人纳入临床试验和 考虑到多国试验中的区域差异。不同的子组似乎 需要不同的方法。也就是说，连续的(例如年龄)、绝对的(例如性别)和重叠 (例如，共病、多药)变量不同，每个变量可能需要不同的统计分析。 会议将讨论传统方法、贝叶斯方法和其他创新模式和 探索每种方法的优点和局限性。此外，可视化和图形化的作用 将讨论代表权问题。并非所有子组分析都必须在 个人临床试验；对批准后的观察数据的分析可能会说明重要的 在产品开发过程中无法发现的子组之间的差异。对于临床医生来说 对于患者来说，重要的因素不是平均治疗效果，而是是否有 产品或干预措施的益处将超过可能服用该产品的个人的风险。

FDA PA 19‐306 Bierer, Barbara E., M.D. Innovative statistical methodologies to subgroup analysis in clinical trials Project Summary The Multi‐Regional Clinical Trial (MRCT) Center of Brigham and Women's Hospital and Harvard (MRCT Center) is a research and policy center created to address the conduct, oversight, ethics and regulatory environment of clinical trials, with a focus on multinational trials. To do the work, we function as a independent convener to engage diverse stakeholders from industry, CROs, academia, patients and patient advocacy groups, non‐profit organizations, and global regulatory agencies to address problems in rigor and integrity of trials. In this proposal, we propose to convene a public conference of statisticians, clinical trialists, regulators, and patient/patient advocates to discuss innovative statistical methodologies to subgroup analysis in clinical trials. This conference is highly relevant to the FDA's efforts to promote inclusion of individuals of diverse backgrounds and characteristics in clinical trials and to considerations of regional differences in multi‐national trials. Different subgroups appear to necessitate different approaches. That is, continuous (e.g. age), categorical (e.g. sex) , and overlapping (e.g. co‐morbidity, polypharmacy) variables differ, and each may command different statistical analyses. The conference will address traditional approaches, Bayesian methods, and other innovative models and explore the advantages and limitations of each. Further, the role of visualization and graphical representation will be discussed. Not all subgroup analyses must be performed at the level of the individual clinical trial; analyses of post‐approval observational data may illuminate important differences across subgroups that were not discoverable during product development. For the clinician and for the patient, the important factor is not the average treatment effect, but rather whether the benefit of the product or intervention will outweigh the risks for the individual likely to take the product.