多水平模型在临床试验中心效应分析中的研究

Multilevel Models in the Analysis of Center Effect in Clinical Trials

多水平模型是既适用于计量资料、又适用于二分类或多分类资料的统计方法。本文通过对各类型的实例数据,分别应用多水平模型、协方差分析和CMH卡方检验方法进行分析,探讨多水平模型在多中心临床试验中心效应分析中的应用。结果表明,在计量资料的中心效应分析中,协方差分析较多水平模型更容易发现中心效应;对分类资料的中心效应,多水平模型更为敏感。可见不同分析方法对中心效应的分析结果不相同,在实际运用中应结合数据自身的结构特点和研究目的,并根据各种方法的适用条件选择分析方法。

Multilevel models are applicable to both the quantitative data and categorical variables. We used the meth- ods, including the multilevel models, analysis of covariance and CMH chi-square test, to analyse different types of data, to explore the application of multilevel models in the analysis of the multicenter clinical trial center effect. The results showed that the analysis of covariance is more sensitive to find the center effect for quantitative data, while multilevel models are more sensitive to categorical variables. It can be seen that results with different analytical meth- ods for center effect are not the same, and the most appropriate method should be selected in accordance with the characteristics of data, the objective of research, and the applicable conditions of the various methods in practicaluse.