Meta分析中若干统计方法的研究及meta分析软件研发

The main purpose of evidence-based medicine is to provide the relatively best and strongest evidence, therefore, the research results of evidence-based medicine have an extremely important and extensive influence on clinical practice, medical research and decision-making in medical care. Most of the research results of evidence-based medicine have been obtained by the meta-analysis. However, there have been some crucial limitations in the methodology and application of meta-analysis: present methods for detecting and correcting publication bias have low power, number of studies is admitted to be too small for conducting meta-analysis with the consequence of potential unreliability, methods of meta-analysis for dealing with some general data types (for instance, meta-analysis of Youden Index) are unavailable at present, and the software for meta-analysis could not meet the application requirements, and so on. Thus these limitations lead to the overestimated combined effect size and incorrect interpretation of the results of meta-analysis, in addition to seriously misleading of clinical practice and medical research and decision-making. Domestic study about the methodology involved in meta-analysis is hardly found in the international publication. This project will carry out a study directed to some key issues involved in meta-analysis, mainly including: to analyze the characteristics which may affect the publication of meta-analysis study, to establish a meta-analysis database, to build up more effective methods for detecting and correcting publication bias, to propose the standard of the minimum number of studies included in meta-analysis, to establish a meta-analysis method for weighted Youden index, and to develop a more powerful professional software of meta-analysis with independent intellectual property rights. The goal of the project is to improve and even overcome the major defects of methodology in the application of meta-analysis with making the evidence obtained from meta-analysis more accurate and reliable.

循证医学以提供相对最好、最强证据为主要目的，故其研究结果对临床实践和医学研究及决策有极其重要和广泛影响。循证医学研究结果绝大多数经meta分析得出，但meta分析在方法学和应用方面有诸多局限：缺乏有效的探测和纠正发表偏倚的方法，没有最小研究数限制，适用的资料类型有限，分析软件不能满足应用要求等，由此导致综合效应量的高估和对meta分析结果的错误解释，严重误导临床实践和医学研究及决策。我国meta分析的方法学研究在国际尚属空白。本项目针对meta分析方法所涉及的若干关键性问题展开研究，刻画meta分析的应用现状和建立meta分析数据库，建立更有效的探测和纠正发表偏倚方法，提出meta分析的最小研究数标准，建立加权Youden指数等meta分析方法，研发功能更强的具有自主知识产权的meta分析专业软件等。最终目的是改善乃至克服meta分析应用中的重大缺陷，使meta分析所得证据更为准确可靠。

循证医学以提供相对最好、最强证据为主要目的，故其研究结果对临床实践和医学研究及决策有极其重要和广泛影响。循证医学研究结果绝大多数经meta分析得出，但meta分析在方法学和应用方面有诸多局限：缺乏有效的探测和纠正发表偏倚的方法，没有最小研究数限制，适用的资料类型有限，分析软件不能满足应用要求等，由此导致综合效应量的高估和对meta分析结果的错误解释，严重误导临床实践和医学研究及决策。我国meta分析的方法学研究在国际尚属空白。本项目针对meta分析方法所涉及的若干关键性问题展开研究，主要完成了一下工作：1) 建立了Meta分析发表文献的重要特征变量数据库，为meta分析的方法学研究奠定了良好基础；2) 进行了异质性检验方法的模拟比较研究，结论为以PM法和EB法普遍估计较为准确；3）建立了新的Youden指数的统计推断方法，纠正了该法应用60余年的错误；4）建立了加权Youden指数的meta分析方法，为此类应用研究提供了新的有效工具；5）对于N-of-1 设计的定量数据分析做了比较研究，meta分析的效果不佳，而混合模型更值得推荐；6) 进行了Meta分析发表偏倚校正的研究，提出了基于Bayesian的发表偏倚校正方法，在研究数大于6时统计性能较好，在方法学上有创新性；7) 进行了发表偏倚对meta分析结果影响的研究，提出了发表偏倚和异质性共同影响综合效应量估计的观点；8) 建立了灵敏度与特异度关联条件下Youden指数的meta分析方法，也是方法学上的创新；9) 建立基于随机效应模型的两标准化均数比较综合效应量的加权方法，初步结果比较乐观；10）研发成功了专用于meta分析的软件METABOX，目前处于著作权最后审理阶段，该软件具有功能强、界面友好、易操作等优势，且为国内首个具有自主知识产权的此类软件。