标题: 贝式理论应用在第二期临床试验结合安全性考虑下使用二阶段方法和第三期临床试验下使用多区域临床试验概念之设计与评估
Bayesian Approaches for design and evaluation of the Phase II clinical trials incorporating safety and the multiregional clinical trials
作者: 鄭宇傑
蕭金福
Cheng, Yu-Chieh
Hsiao, Chin-Fu
統計學研究所
关键字: 二阶段设计;毒性;贝式;先验分配;多区域临床试验;two-stage design;toxicity;Bayesian;prior;MRCT
公开日期: 2017
摘要: 药物开发是一个复杂且需要钜额费用的长时间过程。传统上,这过程需要有三个阶段:寻找最大容忍毒性的剂量,并在这剂量范围内,测试疗效的效果,并寻找是否有其他反映或者副作用。除此之外,为了世界各地患者的权益,全球化临床试验分别在许多不同的区域逐渐地实施。这篇论文主要探讨在第二期临床试验上的二阶段临床试验设计与第三期临床试验的全球化临床试验设计。
第一期临床试验的主要目的是寻找一个新药或者疗程的最大容忍毒性的剂量(MTD)。而第二阶段则是考量在这最大容忍毒性的剂量下,对于预期病症的疗效效果如何。这两个阶段分别讨论毒性与效性。然而,这篇论文使用贝式两阶段方法来设计同时考量毒性与效性下,设计第二期临床试验
为了加速全球化药物开发,越来越多的临床试验需要在世界各地进行。1998年,国际药品法规协和会(ICH)发表了指导手册E5讨论衔接性临床试验,并在2006年的第11届Q&A 文件中讨论多区域临床试验的定义。日本厚生劳动省在2006年发表了一篇指导手册, ‘ Basic Principles on Global Clinical Trials’,对于多区域临床试验中各个区域的疗效,提供了两种方法来如何决定各个区域疗效的一致性。在2016年,国际药品法规协和会发表了新的指导手册E17,更加详细定义与如何实施多区域临床试验。E17建议多区域临床试验不只评估整体疗效效果的效益,且要分析各个区域的疗效情况。
此篇论文主要讨论如何利用贝式方法在设计与评估于第二期临床试验上两阶段设计与第三期临床上多区域临床试验。第一个主题同时考量毒性与效性作为主要反应,并使用狄利克雷作为先验分配。这个方法以毒性的大小来评估疗效的安全。并提供先验分配中毒性与效性的相关性作为参考,来设计阶段上的选择或者评估每个阶段的结果。考量两种情况下,每个阶段以事后机率作为评估是否此试验可以继续进行或者提早终止试验。第二个主题是在多区域的临床试验下,并假设整体疗效效果与区域间变异数皆为随机变数下,来设计与评估此试验。整体效应假设有均匀分布为其先验分配,而区域间变异数则假设有半常态与逆伽玛分配为其先验分配。最后以整体疗效效果的后验分配来做为评估疗效的结果。最后这两个应用皆以实际例子来做示范评估。






关键词: 二阶段设计;毒性;贝式;先验分配;多区域临床试验
Drug development is complicated and expensive with a long life cycle. Traditionally, this process has three phases to find the maximum tolerated dose (MTD); check active efficacy response as a cure, and other response or side effects. Also, global clinical trials are conducting increasingly for benefits of patients in different regions. This thesis discusses two topics related to phase II and III trials, namely two-stage designs and multiregional clinical trial.
The goal of a phase I trial is to find the maximum tolerated dose of a proposed drug or treatment is the main goal in the phase I trial. And the phase II trial proceeds with this dose to evaluate the efficacy of the proposed drug or treatment. These two steps consider toxicity and efficacy, respectively. Therefore, this study discusses Bayesian two-stage designs with considering both efficacy and toxicity as main endpoints simultaneously.
To accelerate global drug development, clinical trials increasingly need to be conducted around the world. Therefore, in 1998, the International Conference on Harmonisation (ICH) published the ICH E5 guideline about bridge study, and the 11th questions-and-answer Q&A document for the ICH E5 in 2006 discusses the definition of a multiregional clinical trial (MRCT). The Japanese Ministry of Health, Labour, and Welfare (MHLW) published an important guideline in 2007, “Basic Principles on Global Clinical Trials”, which provided two methods for determining of the consistency of regional effects. The ICH published in 2016 a new detailed guideline ICH E17 for designing and conducting of an MRCT. ICH E17 recommends that MRCT designers should evaluate not only the overall treatment effect but also the regional treatment effect.
This study constructs and evaluates two-stage designs and MRCT for two applications using Bayesian methods. The first application uses efficacy and safety as main responses, following a Dirichlet distribution as a prior. A phase II trial traditionally only evaluates efficacy. This approach also applies toxicity as main endpoint to evaluate trial safety. The correlation between rates of efficacy and toxicity of priors is influenced by choices made previously in the design stage or evaluation. Sequentially, two scenarios are discussed to evaluate the posterior probabilities at each stage. The MRCT is then designed and evaluated, and the overall treatment effect with given the uniform distribution as prior and between-region variability with given the half-normal or inverse gamma distributions as priors are considered random variables. The posterior probability of the overall treatment effect is then calculated to evaluate efficacy of a treatment. These two applications are then analyzed using two different examples.
key word: two-stage design; toxicity; Bayesian; prior; MRCT
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT079926803
http://hdl.handle.net/11536/140971
显示于类别:Thesis