临床试验计量重复测量资料的统计

目的 探讨药物临床试验计量重复测量资料统计分析方法的选择。方法 通过实例列出各种分析方法并进行比较。结果 临床药物研究资料常为重复测量资料，比较治疗组与对照组测量值是否有显著性差别，在统计方法选择策略上一种是对一个代表性值进行比较，采用传统的统计方法如两独立样本t检验、方差分析和协方差分析；另一种策略是采用混合效应模型对整个研究过程中所有时点的测量值进行分析。结论 两种统计策略均可以采用，但混合效应模型是较好的分析方法。药物临床试验研究资料常是重复测量资料，各时点值之间存在相关性。混合效应模型可以充分利用信息，既可以分析随机效应和相关性，又能处理缺损值。

Data Analysis Strategies for Repeated Measures Data in Clinical Trials

Purpose Clinical trials have received much attention in recent years,but data analyses have concentrated on a single longitudinal variable. This paper compared different statistical methods in SAS pro-cedures for repeated measures in randomized clinical trial and helps you understand their methodologies and advantages. Methods Using an example to illustrate many of the key similarities and difference to various statistical models. Results There are two important data analysis strategies for comparing the difference effect between the two treatment groups. With respect to the two data analysis strategies, statistical methods are classified as traditional methods and statistical methods for repeated measurements'. (1) Traditional statistics include independent t test,analysis of variance and analysis of covariance; (2) Statistical method for repeated measurement includes Generalized Linear Modei and Generalized Linear Mixed Model. A characteri-stic of repeated measurement data in clinical trials is that observation on the same subject are correlated, and the closer the time point,the high correlation. Hence,statistical analysis must address the issue of covariation between measurements on the same subject. The Generalized Linear Modei is not a statistical technique for analyzing correlated outcome data,but the MIXED procedure of the SAS system provides a rich selection. Conclusions Continuous data in clinical trial always involves data consist of multiple measurements on individuals, within individuals data are usually positive correlated. Both strategies offers analyses that account for between group differences,but most situations, the Generalized Linear Mixed Modei is the most preferable for repeated measurement data in clinical trial. It covers both random effects and serial correlation and allows for missing data.