网状meta分析实效性评价方法及其广义线性混合效应模型的构建与应用研究

项目名称： 网状meta分析实效性评价方法及其广义线性混合效应模型的构建与应用研究

项目编号： No.81502896

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 医药、卫生

项目作者： 于莉莉

作者单位： 中国人民解放军联勤保障部队第九六〇医院

项目金额： 18万元

中文摘要： 组间直接比较证据缺乏的情况下，采用网状Meta分析对多种干预措施进行比较是近年来新兴的一门循证统计学技术。它将所有直接和间接证据集成于一个总的统计框架中，成为Meta分析方法学最重要的一个发展方向。然而，由于该方法部分基于间接证据，证据传递导致信息损失或偏性，故其理论完备性与实效性一直备受争议和质疑，如数据间的同质性假设、相似性假设及一致性假设的判定方法及关联程度，弱关联效应及偏倚的评估，统计合并模型的选择及其适用条件等。这些问题的存在，使网状Meta分析纵然有引人注目的网络结构图形，也只是表象上的证据，存在着较高的证据假象的风险。本课题旨在利用计算机模拟技术与临床研究实例数据演算方法，拟从方法学假设、网络设置、偏倚与证据质量评估、统计分析模型等方面探讨网状Meta分析实效性评价方法，进而构建网状Meta分析广义线性混合模型以探讨其实效性，研究结果有望为临床决策提供方法支持与实践指导。

中文关键词： 统计方法；临床试验；网状Meta分析；统计模型

英文摘要： The network meta-analysis is introduced as a new popular and powerful statistical technique used in the evidence-based medicine that allow for comparisons of two or more different interventions that have not been studied in a head to head fashion. It incorporates all available direct and indirect evidence into a general statistical framework and became one most important recent development in meta-analysis methodology. Unfortunately, due to the use of indirect evidence and the mutual transitivity of evidence in network meta-analysis, the validity of this method are still controversial and questionable. For examples, What about the relationship among the assumptions of homogeneity, similiarity and consistency and how to judge whether the three assumptions are satisfied? How to estimate the weak correlation effect and risk of bias? What about the precision of different analysis model and the conditions they required? These issues will make the network meta-analysis artificially appear to provide a stronger evidence base accompanying graphical representation of a treatment network but potentially false impression of the strong evidence base. In this subject,we wants to explore the methods for assessing the validity of network meta-analysis in method assumptions, treatment network, risk of bias assessment and evidence synthesis models, then we will investigate a fixed-effect or random-effect generalized linear mixed models for network meta-analysis and evaluate it’s power for assessing the validity of network meta-analysis. All these above work will be performed depending on computer simulation and real clinical research examples. The purpose of this project is to provide constructive consultation in practicing network meta-analysis in clinical research and help clinical decision-making.

英文关键词： Statistical Method;Clinical Trial;Network Meta-analysis;Statitical Model