Abstract–To evaluate a novel vaccine, randomized clinical trials are often conducted to assess how much the infection or disease rate is reduced in the vaccinated group as compared to the unvaccinated group. Because of low incidence rate for the clinical endpoint, the sample size based on exact conditional binomial test is often very large for vaccine efficacy trials, which poses big challenge for study conduct and enrollment, especially in studies with very low incidence rates. Bayesian framework provides a natural avenue to use historical information, if there is evidence to suggest similarity of the responses between the historical and current studies. In this article, we first propose a hierarchical conditional binomial model to provide a Bayesian framework to use historical information for assessing vaccine efficacy. Secondly, we propose a beta prior distribution which is equivalent to the power prior to borrow partial historical information quantitatively. Simulations are conducted to evaluate the power and Type I error for vaccine efficacy study by borrowing historical information with the proposed analytic approaches. The proposed methods are demonstrated by an example to show the improved efficacy estimation through borrowing information from a historical study.

摘要-为了评估新型疫苗，经常进行随机临床试验以评估与未接种疫苗的组相比，接种疫苗组的感染或疾病发生率降低了多少。由于临床终点的发生率较低，因此基于精确条件二项式检验的样本量对于疫苗功效试验而言通常非常大，这对研究进行和招募构成了巨大挑战，尤其是在发生率极低的研究中。如果有证据表明历史研究与当前研究之间的相似性，贝叶斯框架提供了使用历史信息的自然途径。在本文中，我们首先提出一个分层条件二项式模型，以提供一个贝叶斯框架来使用历史信息来评估疫苗效力。其次，我们提出一个beta先验分布，它等同于定量借用部分历史信息之前的幂。通过使用所提出的分析方法来借鉴历史信息，进行了仿真以评估疫苗功效研究的功效和I型误差。通过实例演示了所提出的方法，以通过借鉴历史研究中的信息来显示改进的功效估计。