Response adaptive randomized clinical trials have gained popularity due to their flexibility for adjusting design components, including arm allocation probabilities, at any point in the trial according to the intermediate results. In the Bayesian framework, allocation probabilities to different treatment arms are commonly defined as functionals of the posterior distributions of parameters of the outcome distribution for each treatment. In a non-conjugate model, however, repeated updates of the posterior distribution can be computationally intensive. In this article, we propose an adaptation of sequential Monte Carlo for efficiently updating the posterior distribution of parameters as new outcomes are observed in a general adaptive trial design. An efficient computational tool facilitates implementation of more flexible designs with more frequent interim looks that can in turn reduce the required sample size and expected number of failures in clinical trials. Moreover, more complex statistical models that reflect realistic modeling assumptions can be used for analysis of trial results.

中文翻译：

---

顺序蒙特卡洛用于适应性随机试验。

响应适应性随机临床试验由于可以灵活地根据中间结果在试验的任何时候调整设计组成部分（包括臂分配概率）而变得越来越受欢迎。在贝叶斯框架中，分配给不同治疗组的概率通常定义为每种治疗的结果分布参数的后验分布函数。但是，在非共轭模型中，后验分布的重复更新可能需要大量计算。在本文中，我们提出了一种适应性顺序蒙特卡洛方法，用于在常规适应性试验设计中观察到新结果时有效地更新参数的后验分布。高效的计算工具有助于以更频繁的临时外观实施更灵活的设计，从而可以减少所需的样本量和预期的临床试验失败次数。而且，可以使用反映实际建模假设的更复杂的统计模型来分析试验结果。