提高国内随机对照试验Meta-分析的质量

高质量的随机对照试验(RCT)的Meta-分析结果与国际公认的大样本RCT结果一起被各国列为最高等级的证据,可为临床实践和卫生决策提供更真实的科学依据,引导临床医师在实践中做出正确的决策。但质量差的Meta-分析反而可能导致错误的结论。国外Meta-分析方法的应用已趋于成熟和规范。然而目前国内由于应用时间不长,且缺乏统一规范的实施标准,文献质量参差不齐,很多方面还存在着较为严重的缺陷。探讨如何提高国内RCTMeta-分析的质量,尽快与国际接轨,以便为循证医学、药学提供更科学真实的证据。以进行RCTMeta-分析的步骤为线索,对国内Meta-分析存在的问题和解决的办法进行了综述。具体内容包括:提出一个好问题(研究目的),全面收集相关的RCT,制定严格的纳入/排除标准,正确提取数据资料,对符合纳入标准的RCT进行质量评价,应用正确的统计方法,必须进行敏感性分析,根据结果做出正确、全面的结论。一篇理想的RCTMeta-分析应当纳入当前所有的高质量的同质研究,无发表偏倚,使用正确的模型和统计学方法,并对结果做出全面的论述。也就是说只有尽量满足进行Meta-分析每一步所需的条件,才能逐步提高Meta-分析的质量,使得其结论更加科学,可靠。

Improve the qualities of meta-analyses of randomized controlled trials in China

The meta-analytic results of the hig h qualitative randomized controlle d trails(RCT)and the results of the large sampling RCT that is acknowledged in the world are recognized to be the most au thoratative evidences.Those evidences can provide more reliable and scient ific foundation for clinical practi ce and health policy-making.Moreover,th e clinical doctors can use these evid ences to make the best choice during their practice.On the contrary,meta-analysis with poor quality may result in error conclusion.The meta-analysis is be-coming more and more mature abroad.However,in our country there still are several serious defects of meta-analysis.To discuss how to improve t he qualities of meta-analyses of randomized controlled trials in China in o rder to provide more reliable and scientific evidences for EBM and pharmacology.According to procedures of the meta-analysis of RCT,we review the prob-lems and their resolutions of meta-a nalyses methods that are employed wi thin China,concluding:questioning(the purpose of the research),comprehen-sive collecting relative RCT,formu lating strict standard which decides if the patients can be bring into or removed,correctly getting out the data,as-sessment of RCT,using correct stati stic methods,analyzing sensitivit y,making comprehensive and correct conclusion.The ideal meta-analysis o f RCT must conclude all high qualitative RCT currently,no public bias,using correct model and statistic methods and comprehensive explanation of th e conclusion.In a word,the more condi tions that are satisfied in each step of the analysis,the higher quality results we get.