含治愈个体生存资料的亚组识别研究

Study of Subgroup Identification for Failure Time Data with a Cured Fraction

目的基于K指数构建一个用于检验生存资料是否存在亚组的统计量,若存在亚组则利用亚组相关协变量建立亚组判别模型预测病人的亚组身份。方法基于AFT模型与AFTMC模型所得K指数构建统计量K sub检验生存资料是否存在亚组,当存在亚组时构建统计量K off寻找区分亚组的最佳时间点,并确定每位患者的亚组身份,进而建立亚组判别模型。结果K sub的Ⅰ类错误基本控制在0.05以内,检验效能在多数情况下能够保持较高水平,但当样本量较少、治愈率低以及删失率较高时,K sub的检验效能下降。在用K off寻找到区分亚组的最佳时间点T(off)之后,在自定义的四种方法中,方法三识别患者亚组身份的平均灵敏度、特异度和准确度分别为86.8%、82.5%和89.7%,标准差分别为4.1%,5.3%和6.7%。在与自定义的四种方法相对应的四个判别模型中,模型三预测新入组患者亚组身份的平均灵敏度、特异度、准确度和AUC均最高(分别为93.1%、77.5%、82.7%和87.6%),波动范围均最小(标准差分别为6.7%、7.2%、5.1%和3.3%)。结论生存资料可用统计量K sub检验是否存在亚组。若存在亚组,方法三能够准确和稳定地识别患者的亚组身份,模型三能够有效地预测新入组患者的亚组身份。

Objective Based on the K index,we intend to construct a statistic to test whether there is a subgroup in survival data.If there is a subgroup,we establish a subgroup discriminant model to determine the subgroup identity of each patient.Methods Based on the K index of the AFT model and AFTMC model,we construct a statistic of K sub to test whether there is a subgroup in survival data.When there is a subgroup,we construct the statistic of K off to find the optimal time point for distinguishing subgroups,determine the subgroup identity of each patient,and then establish a subgroup discriminant model.Results The type I error of K sub is basically controlled within 0.05.The power can be maintained at a high level in most cases,but when the sample size is small,the cured rate is low,and the censored rate is high,the power of K sub declines.After finding the optimal time point(T(off))for distinguishing subgroups by K off,the average sensitivity,specificity and accuracy that Method 3 identify patient subgroup identities are 86.8%,82.5%and 89.7%,the standard deviation are 4.1%,5.3%and 6.7%.In the four discriminant models corresponding to the four customized methods,the average sensitivity,specificity,accuracy and AUC that Model 3 predictsubgroup identity of newly enrolled patients are the highest(93.1%,77.5%,82.7%,and 87.6%,respectively),and the fluctuation range are the smallest(standard deviation is 6.7%,7.2%,5.1%and 3.3%,respectively).Conclusion K sub can be used to test whether there is a subgroup in survival data.If there is a subgroup,Method 3 can accurately and stably identify the subgroup identity of the patient,and Model 3 can effectively predict the subgroup identity of the newly enrolled patient.