【作者】 周涛；

【导师】 孙奕；

【作者基本信息】 华中科技大学 ， 流行病与卫生统计学， 2013， 硕士

【摘要】 目的：对我国医学期刊中发表的Meta分析文章质量进行评价，重点对统计方法学方面质量问题进行评价。方法：以2011年国内中文期刊发表的有关药物治疗类的Meta分析文章作为评价对象，制定标准筛选出符合标准的文章，用自行编制的《Meta分析质量评价表》对纳入的Meta分析文章进行统计方法学方面的质量问题进行评价，评价内容包括四个方面：异质性检验、数据合并、发表偏倚控制、敏感性分析。同时采用OQAQ量表对选出的Meta分析文章整体质量进行评价，并探讨统计方法学方面质量评价结果与QQAQ量表评价结果的关系。用频数和构成比描述Meta分析文章的作者单位类型、研究类型、资金赞助类型、文献纳入数量等基本情况，对方法学质量状况进行分类统计；采用单因素分析和多重对应分析等方法分析影响Meta分析文章质量的有关因素。结果：本次研究共纳入214篇文献。OQAQ量表平均得分为5.67±1.18（总分为7分）；总的合格率为27.10%，不合格率最高的三个条目分别为检索策略（38.97%不合格）、选择偏倚的处理（37.38%不合格）以及数据合并（35.05%不合格）；完全符合《Meta分析质量评价表》评价条目的文章仅有9篇，总合格率仅为4.21%。纳入文献的敏感性分析、发表偏倚、数据合并方面存在的主要问题表现为：有166篇（77.57%）未进行敏感性分析，140篇（65.36%）未进行发表偏倚检验，有75篇（35.05%）的文章存在数据合并不合理现象。多重对应分析结果显示Meta分析文章质量问题与作者的资金赞助、入选文献篇数、作者类型存在关联。结论：本次研究发现中文医学期刊发表的Meta分析文章在方法学质量方面普遍存在问题，特别是在抽样偏倚和选择偏倚的控制、数据合并、发表偏倚检验和敏感性分析这几个方面尤为突出。创新点：本次研究首次结合OQAQ量表和《Meta分析质量评价表》评价中文期刊Meta分析文章的质量。更多还原

【Abstract】 Objective:To evaluate quality of statistical methods and methodological quality of Meta-analysisliteratures published in domestic medical journals.Methods:Domestic Meta-analysis literatures on drug treatment of year2011are selected in thisstudy. Self-designed evaluation scale of methodological quality of meta-analysis andOQAQ scale are used to evaluate the quality of methodology and statistical analysis.Statistical methods mainly involve four aspects: test of heterogeneity, data combining,publication bias analysis, sensitivity analysis. The frequency and constituent ratio are usedto describe general situation of literatures (type of study, author affiliation, type ofsponsored scholarship and number of selected articles) and quality of methodology.Describe the quality of four aspects involved in statistical analysis according to differentgeneral characteristics of literatures. Multiple correspondence analysis is used to revel thecorrespondence relationship between the general characteristics and the quality of statisticalanalysis.Result： 214literatures are included in this study, mean score of OQAQ scale of which is5.67±1.18. Three major problems exist in these literatures:38.97%of literatures’ areunqualified in search strategy,37.38%are unqualified in dealing with selection bias and35.05%are unqualified in data combining. There are only21literatures that all meet thefirst9standards of OQAQ scale. Result of evaluation scale o f statistical analysis ofmeta-analysis shows that only9literatures are totally qualified, and the main problemsexist in sensitivity analysis, publication bias analysis and data combining. Sensitivityanalysis isn’t conducted in166(77.57%) literatures. publication bias isn’t explored in140literatures65.36%).data combining method is unreasonable in75(35.05%) literatures.Result of multiple correspondence analysis indicates that methodological quality aresignificantly associated with scholarship of funds, number of selected articles and type ofauthor.Conclusion:This study indicates that published meta-analysis articles in domestic medical journalshave defects in methodology, especially in statistical methods. Of all the defects, thefollowing ones are more prominent: common existence of sampling bias and selection bias,unreasonable consolidation, ignorance of test of publication bias and sensitivity analysis.Innovation:This study assessed the quality of the meta-analysis articles in Chinese medicaljournals firstly by the combination of the OQAQ scale and the quality evaluation scale formeta-analysis.更多还原