应用漏斗图识别发表性偏倚的效率研究
Study on Effectiveness of Funnel Plot Used in Detecting Publication Bias
杨书 1李婷婷 2刘新1
作者信息
- 1. 成都医学院,公共卫生学教研室,成都,610083
- 2. 药物分析教研室,成都,610083
- 折叠
摘要
目的 在Meta分析中,发表性偏倚的存在常常会导致对效应量的过高估计,漏斗图是目前较常用的用以识别发表性偏倚的方法之一,本研究通过对发表性偏倚的可能性来源与漏斗图工作原理的深入探讨,对漏斗图判断发表性偏倚的效率进行评价.方法 采用问卷调查的形式,计算准确度(AC)、正确指数(Youden's in-dex)等指标.结果 实验组A、B、c三型问卷的平均准确度分别为52.3%、57.5%、62.2%,对照组分别为51.2%、54.6%、60.1%.两组比较差别无统计学意义.结论 用漏斗图方法来识别发表性偏倚效率低且易得出错误结论.
Abstract
Objective Publication bias can lead to overestimation in Meta - analysis.The funnel plot is frequently used to detect such biases as a usual tool. The aim of this article is to determine whether the researchers can correctly identify publication bias from visual inspection of funnel plots in diverse - size Meta - analysis. Methods The Youden's index and the percentage of plots which were correctly identified were calculated from the questionnaires. Results The mean percentages of plots correctly identified by experiment group in the three type questionnaire were 52.3% , 57.5%., 62.2% respectively.These data were not significantly different from the controls which were 51.2%, 54.6%, 60.1% respectively.Conclusion The effectiveness of funnel plot used in detecting publication bias was low and this method often led to mistake conclusion.
关键词
漏斗图/发表性偏倚/Meta-分析/问卷Key words
funnel plot/publication bias/Meta - analysis/questionnaire引用本文复制引用
出版年
2007
NETL
NSTL
维普
万方数据