Meta回归与亚组分析在异质性处理中的应用  

Application of Meta-regression and subgroup analyses of heterogeneity disposal in Meta-analysis

作　　者：石修权[1] 王增珍[1] 

机构地区：[1]华中科技大学同济医学院公共卫生学院流行病与卫生统计学系,武汉430030

出　　处：《中华流行病学杂志》2008年第5期497-501,共5页Chinese Journal of Epidemiology

基　　金：国家“973”重点基础发展规划资助项目（2002CB512910）

摘　　要：探讨Meta回归与亚组分析在异质性的识别与处理中的应用及意义。利用文献提供的二次数据建立Meta回归模型，筛选出异质性的影响因素，根据该因素做亚组分析，并比较亚组分析前后异质性的变化。Meta分析资料经异质性检验，Q=44．71，df=27，P=0．017，认为存在异质性。经Meta回归分析，从可能导致异质性的因素（研究时间、地区、样本量、病例对照比值等）中筛选出样本含量为异质性因素（P=0．012）、地区为可能的异质性因素（P=0．091）。然后进行亚组分析，异质性明显减小（∑Q由44．71减小至32．11）。结论：Meta回归法对筛选异质性影响因素比较简便可靠，据此进行的亚组分析能明显降低亚组内的异质性。故存在统计学异质性又要计算合并效应时推荐二者结合使用，可正确识别并降低异质性，从而使Meta分析结果更为稳健与合理。To explore the role and application of Meta-regression and subgroup analyses to recognize and control the heterogeneity in Meta-analysis, Meta-regression models were established by secondary data to screen the factors resulting heterogeneity, and subgroup analyses were used to compare the change of heterogeneity before and after. The heterogeneity was found in the Meta-analysis （Q = 44.71,df= 27, P 0.017）. Sample size and region were selected （P= 0.012 and P= 0.091, respectively） by Meta-regression from many possible factors such as sample size, year, region and case/ control ratio. The Q values were lowered from 44.71 to 32.11 after subgroup analyses. Thus, Metaregression method was convenient and reliable to screen the affected factors of heterogeneity, and subgroup analyses based on the hypothesis that could significantly lower the heterogeneity. It was recommended to a combined use when an obvious heterogeneity existed but was in need to get an overall result in Meta- analysis. We could correctly judge and lower the heterogeneity to increase the robustness and rationality of results from Meta-analysis.

关 键 词：META分析 异质性 Meta回归 亚组分析 

分 类 号：R195[医药卫生—卫生统计学]