配对设计生存资料的统计分析及其在临床试验中的应用

Statistical Analysis of Paired Survival Data and its Application to Clinical Trial

目的 介绍配对设计生存资料的统计分析方法及其在临床试验中的应用。方法 运用基于协方差矩阵稳健估计的Cox回归模型 ,通过调用STATA统计软件包中stcox命令中的cluster选项 ,对配对设计生存资料进行统计分析。并用治疗烧伤临床试验中的实例比较了运用和不运用配对设计生存资料的统计分析方法的两种结果。结果 实例分析表明运用配对设计生存资料的统计分析方法时 ,可得出治疗组的创面愈合时间小于安慰剂组的结论 ,而不运用配对设计生存资料的统计分析方法时 ,其结论为治疗组和安慰剂组创面愈合时间的差异无统计学意义。结论 在对配对设计的生存资料进行统计分析时 ,应选用配对设计生存资料的统计分析方法 ,否则将降低检验的效率 。

Objective Statistical methods for paired survival analysis were introduced and applied to clinical trial in the paper.Methods Cox regression model based on robust variance-covariance matrix estimation was used in analyzing paired survival data through cluster option of stcox command in STATA software. Difference between results of paired survival analysis and ordinary survival analysis were explained through an example of clinical trial of curing burns.Results Statistical significance were observed in difference between times of cicatrizing of treat group and that of placebo group while paired survival analysis was used. Yet using ordinary survival analysis, no statistical significance were observed between the two groups.Conclusion In order to get correct answers, paired survival analysis should be used for paired design survival data. Or hypothesis test efficiency would be much lower and wrong conclusion would be resulted.