Utility-Based Dose Selection for Phase II Dose-Finding Studies
Utility-Based Dose Selection for Phase II Dose-Finding Studies
复制标题
DOI：
10.1007/s43441-021-00273-0
复制
发表时间：
2021-04-13
影响因子：
1.5
通讯作者：
Darchy, Loic
中科院分区：
文献类型：
--
作者：
Aouni, Jihane;Bacro, Jean Noel;Darchy, Loic
背景和目的剂量选择是临床发展的一个关键特征。不良的剂量选择已被认为是开发后期失败的主要驱动因素。它通常包括有效性和安全性标准。本文的目标是开发和实施一种新的全贝叶斯统计框架，通过最大化第三阶段的预期效用来优化剂量选择过程。方法采用疗效和安全性两部分的效用函数来表征成功率。每个成分都是一个剂量-反应模型。此外，顺序设计（在中期分析中具有无效和有效性规则）与固定设计进行比较，以便允许人们加快执行后期研究的决定。在大范围的剂量-反应情景下，通过模拟广泛地评估了这种方法的工作特性。模拟结果表明，很难同时估计两个复杂的剂量-反应模型，并使其具有足够的精度，从而使用结合两者的效用函数对剂量进行适当排序。做出正确决策的概率随着样本量的增加而增加。在某些情况下，顺序设计具有良好的特性：在中期分析时研究终止的概率相当大，因此可以在保持固定设计特性的同时减少样本量。
Background and Objectives Dose selection is a key feature of clinical development. Poor dose selection has been recognized as a major driver of development failure in late phase. It usually involves both efficacy and safety criteria. The objective of this paper is to develop and implement a novel fully Bayesian statistical framework to optimize the dose selection process by maximizing the expected utility in phase III. Methods The success probability is characterized by means of a utility function with two components, one for efficacy and one for safety. Each component refers to a dose-response model. Moreover, a sequential design (with futility and efficacy rules at the interim analysis) is compared to a fixed design in order to allow one to hasten the decision to perform the late phase study. Operating characteristics of this approach are extensively assessed by simulations under a wide range of dose-response scenarios. Results and Conclusions Simulation results illustrate the difficulty of simultaneously estimating two complex dose-response models with enough accuracy to properly rank doses using an utility function combining the two. The probability of making the good decision increases with the sample size. For some scenarios, the sequential design has good properties: with a quite large probability of study termination at interim analysis, it enables to reduce the sample size while maintaining the properties of the fixed design.