定性临床试验资料meta分析的经验贝叶斯模型原理和应用  

Empirical Bayesian Hierarchical Model for Meta Analysis with Application to Qualitative Clinic Trial Data

作　　者：曾平[1] 王婷[1] 黄水平[1] 赵华硕[1] 

机构地区：[1]徐州医学院公共卫生学院,221002

出　　处：《中国卫生统计》2012年第5期657-660,共4页Chinese Journal of Health Statistics

基　　金：徐州医学院课题资助;江苏省教育厅高校哲学社会科学研究基金资助项(2010SJB790037);徐州医学院公共卫生学院课题资助(201009;201107;201115)

摘　　要：目的研究meta分析经验贝叶斯分层模型的原理及其在定性临床试验资料分析中的应用。方法以一个激素预防新生儿肺透明膜病的临床试验数据为例,建立分层模型,采用经验贝叶斯的方法进行参数估计和后验推断,尝试应用三种不同的方法估计分层模型中的超参数ν。结果分层模型的经验贝叶斯分析结果表明临床使用激素能够降低新生儿肺透明膜病的发生。结论相对meta分析的随机效应模型,经验贝叶斯分层模型提供了更灵活的分析策略。Objective To study the method of Empirical Bayesian hierarchical model for meta analysis and apply to qualitative clinic trial data. Methods Hierarchical meta analysis was applied to a clinic tri- al data about prevention of hyaline membrane disease using hormone, and parameter estimations and posterior inference were implemented based on empirical Bayes. Three methods were proposed to estimate the hyperparam- eter v. Results Empirical Bayesian analysis implied that using hormone could significantly decrease the occurrence of hyaline membrane disease.Conclusion Empirical Bayesian hierarchical model provides a more flexible framework for meta analysis compared to random effects model.

关 键 词：META分析 分层模型 经验贝叶斯 超参数 新生儿肺透明膜病 

分 类 号：O212.8[理学—概率论与数理统计]