Cox比例风险模型与加速失效时间模型在基因表达数据生存分析中的应用及比较∗

摘要

目的：：探讨Cox比例风险模型与加速失效时间模型( accelerated failure time model,AFT)在基因表达数据生存分析中的应用及比较。方法：针对基因表达数据高维小样本的特性,首先采用偏最小二乘法对基因数据集进行降维,然后以降维后的成分为协变量对两类生存模型进行拟合并比较其性能。结果：两类模型中病人生存时间的对数秩检验表明,不同风险组生存时间的差异均有统计学意义( P <0.01),而AFT模型比Cox模型具有更大的检验统计量的值。结论：Cox模型和AFT模型都适用于基因表达数据的生存分析,在某些实际应用中AFT模型的拟合效果可能更优于Cox模型。%Objective:To explore the application and comparison of Cox proportional hazard ( Cox PH) model and accelerated failure time ( AFT) model in gene expression data survival analysis. Methods:Partial least square method was adopted to reduce the dimensionality of gene dataset on the basis of the high dimension and small sample of gene expression data. In the use of the main componets as covariates, fitting of the two survival models was made and their fitting effects were compared. Results:Log rank tests of the survival time of the patients in these two models indicated that there was significant difference in the survival time of different risk groups ( P <0. 01). The test statistic value in AFT model was greater than that in the Cox PH model. Conclusion:Both of the models are appropriate for survival analysis of gene expression data. In some practi-cal applications, the fitting effect of AFT model may be better than that of Cox PH model.