中国循证儿科杂志 ›› 2019, Vol. 14 ›› Issue (3): 205-211.DOI: 10.3969/j.issn.1673-5501.2019.03.009

• 论著 • 上一篇    下一篇

利用SAS软件PROC MCMC过程步实现诊断性试验的贝叶斯Meta分析

许佩华, 郑建清   

  1. 福建医科大学附属第二医院放射治疗科 泉州,362000
  • 收稿日期:2019-04-25 出版日期:2019-06-25
  • 通讯作者: 郑建清,E-mail:18060108268@189.cn
  • 基金资助:
    福建医科大学附属第二医院苗圃基金:2017MP04

Bayesian meta-analysis of diagnostic tests using PROC MCMC program in SAS software

XU Pei-hua, ZHENG Jian-qing   

  1. Department of Radiation Oncology, The Second Affiliated Hospital of Fujian Medical University, Quanzhou 362000, China
  • Received:2019-04-25 Online:2019-06-25
  • Contact: ZHENG Jian-qing, E-mail: 18060108268@189.cn

摘要: 目的 介绍利用SAS软件PROC MCMC过程步实现诊断性试验的贝叶斯Meta分析。方法 利用Menke编制的基于SAS PROC MCMC过程步的Meta分析代码,以高分辨率超声诊断颞下颌关节盘前移位准确度的研究作为示例,介绍诊断性试验的贝叶斯Meta分析实施方法。结果 基于实例数据,本文给出了贝叶斯固定效应模型和随机效应模型Meta分析结果。随机效应分析结果显示:高分辨率超声诊断可复性盘前移位的合并敏感度0.852(95% CI:0.766~0.924),合并特异度0.865(95%CI=0.760:0.944),诊断比值比DOR 53.413(95% CI:11.855~170.4)。结论 SAS PROC MCMC是实现敏感性和特异性的双变量随机效应贝叶斯Meta分析的优秀方法。

关键词: 贝叶斯Meta分析, 分层模型, 马尔科夫链蒙特卡洛方法, 系统评价, 诊断性试验

Abstract: Objective To introduce the Bayesian meta-analysis of the diagnostic test using the SAC software PROC MCMC program.Methods Meta-analysis code based on SAS PROC MCMC program prepared by Menke J was used and the study on the accuracy of high-resolution ultrasound diagnosis of anterior disc displacement of temporomandibular joint published by Dong Xiaoyu et al. was selected as an example data. Bayesian meta-analysis implementation method for diagnostic test was introduced.Results Based on the example data, the Results of Bayesian fixed effect model and random effect model meta-analysis were given. The Results of random effects analysis showed that the combined sensitivity, specificity, diagnostic Odds Ratio (DOR)of high-resolution ultrasound diagnosis for anterior disc displacement with reduction (ADDWR) were 0.852 (95%CI: 0.766-0.924), 0.865 (95%CI: 0.760-0.944), 53.413 (95%CI: 11.855-170.4) respectively.Conclusion SAS PROC MCMC was an excellent method for achieving bivariate random effects Bayesian meta-analysis of sensitivity and specificity.

Key words: Bayesian meta-analysis, Diagnostic test, Hierarchical models, Markov chain Monte Carlo methods, Systematic reviews