文献速递流行病学与生物统计临床试验与 GCP

III期试验中贝叶斯分析应用不足且报告质量欠佳:一项系统综述

Bayesian analysis from phase III trials was underused and poorly reported: a systematic review.
PubMed 文献 · J Clin Epidemiol · 2020-04-04

该系统综述检索Cochrane CENTRAL、Ovid MEDLINE和EMBASE数据库,纳入49篇使用贝叶斯方法分析主要结局的III期试验报告,评估其遵循ROBUST和BayesWatch指南的报告情况。结果显示贝叶斯方法在III期试验中应用不足,且关键方法学组件(如先验分布)的报告常不完整。

要点

  • III期试验中贝叶斯方法应用不足,报告常缺失先验分布等关键方法学要素。
  • ROBUST指南正确报告条目中位数为7条中的4条,BayesWatch为10条中的7条。
  • 仅22%的研究对先验分布给出理由,57%报告了敏感性分析。

核心结论

  • 贝叶斯方法在III期试验中应用不足,报告常缺乏先验分布等关键方法学组件。

研究设计与样本

  • 系统综述检索Cochrane CENTRAL、Ovid MEDLINE和EMBASE数据库,纳入49篇使用贝叶斯方法分析主要结局的III期试验报告。

主要发现

  • ROBUST指南正确报告条目中位数为7条中的4条(IQR 3-5),BayesWatch为10条中的7条(IQR 6-8)。
  • BayesWatch中最常报告的条目最不具体(目的、设计和结果报告),而具体贝叶斯条目(分析技术和先验设定)报告较差。
  • 三分之二的研究给出了先验分布,22%对先验分布给出了理由,57%的试验报告了敏感性分析。

临床怎么用

  • 解读III期贝叶斯试验报告时,应关注先验分布及其理由、敏感性分析等关键方法学信息的完整性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

OBJECTIVES: We assessed the reporting of replicability items with the BayesWatch guidelines and crucial items of Bayesian analyses with the Reporting Of Bayes Used in clinical STudies (ROBUST) guidelines in reports of phase III trials. STUDY DESIGN AND SETTING: A literature search of Cochrane CENTRAL, Ovid MEDLINE, and EMBASE database was conducted to identify phase III trials using Bayesian methods for their primary outcome. Two reviewers performed the study selection process independently. RESULTS: From 5,404 articles retrieved, 49 reports of trials were included. The median [interquartile range] number of correctly reported items was four of seven [3-5] for ROBUST guidelines and seven of 10 [6-8] for BayesWatch guidelines. The proportion of correct reporting widely varied among items. The most frequently reported items from BayesWatch were the least specific (objectives, designs, and reporting of results). Specific Bayesian items (analytic technique and prior specifications) were poorly reported. Priors were given for two-thirds of studies and were justified for 22%. Sensitivity analyses were reported for 57% of trials. CONCLUSION: Bayesian methods are underused in phase III trials. Reports frequently lack descriptions of essential methodological components of the Bayesian analysis, such as the description of priors, which remains a critical key to interpreting Bayesian analyses. PROSPERO: CRD42017054929.

来源:J Clin Epidemiol,查看原文。版权归原作者所有。

引用

  1. Ferreira D, Vivot A, Diemunsch P, et al. Bayesian analysis from phase III trials was underused and poorly reported: a systematic review. J Clin Epidemiol. 2020. PMID: 32259583. doi:10.1016/j.jclinepi.2020.03.021

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。