贝叶斯分层模型在医疗器械临床试验中的应用

贝叶斯方法是基于贝叶斯定理而发展起来的,用于系统阐述和解决统计问题的方法.贝叶斯方法的核心在于参数随机化,在先验概率的基础上通过参数的后验概率进行统计推断.医疗器械往往具备优良的先验信息,贝叶斯方法在器械临床试验中的应用贯穿试验设计和数据分析的各个阶段,贝叶斯方法在正确应用的前提下,临床试验的成本会比频率学派更小.贝叶斯分层模型与经典贝叶斯方法相比,对先验信息的可交换性要求更低,更为灵活的借取“部分”先验信息.本文以一项冠脉支架临床试验为例,应用贝叶斯分层模型方法,对实际结果与传统频率学派方法获得的结果进行比较,并进行相应的讨论.

Application of Bayesian hierarchical model in clinical trial of medical device

Bayesian method was derived from Bayes theorem.The estimated parameter was a random variable in Bayesian method.Statistical inference was based on the prior probability and obtained from the posterior probability.Information of the prior was commonly well established in medical device study.The cost of study could be minimized if the Bayesian method was used appropriately.Compared to the traditional Bayesian method,Bayesian hierarchical model had less restriction on the interchangeability of the prior.The prior would be partly draw and then combine with the current data.The data from a registry study of drug eluting stent were used as an example,we would show the difference between Bayesian and frequency methods and give more discussion.