多终点临床试验中多重性问题的统计学方法介绍与应用案例

Statistical methods and application cases examples for multiplicity issues in multiple endpoints clinical trials

摘要

多次检验不调整检验水平而导致总Ⅰ类错误率(familywise error rate, FWER)远远高于预期的情况称为多重性问题。本文首先将基于临床终点的分类和特征介绍多重性问题的产生机制。其次，介绍解决多重性问题的统计检验策略及方法，包括平行策略、序贯策略及其组合策略，不同检验策略的检验结果会有所差异。最后将以国内外研究者发起的临床试验为案例，详述上述常见策略方法的实际应用。临床试验中多终点的多重性调整可通过单一策略或多种策略组合实现。根据所选策略或组合，确定每个检验假设的统计分析方法和显著性水平α,以有效控制多重性问题。

Abstract

Performing multiple tests without adjusting the test level results in a higher-than-intended overall familywise error rate (FWER). This phenomenon is known as the multiplicity problem. In this paper, we  first introduced the mechanism of multiplicity problem based on the classification and characterization of clinical endpoints. Then,  strategies and methods to solve the multiplicity problem were introduced, including the parallel strategy/single-step method, the sequential strategy/multistep method/stepwise method, and their combinations. The results of different analytic strategies may vary. The practical application of the above common strategies and statistical methods were introduced through the case studies of domestic and foreign investigator initiated clinical trial. Multiplicity adjustment for multiple endpoints in clinical trials can be achieved by a single strategy or a combination of strategies. Depending on the selected strategy or combination, the statistical  methods and significance level (denoted as α) for each test hypothesis are determined to effectively control the multiplicity problem.

关键词

临床试验 / 多终点 / 多重性 / 检验策略 / 统计学方法

基金资助

北京市自然科学基金项目