微信直播

AME循证杂谈023|有关生存资料预后研究样本量计算(Logrank Test)——基于PASS软件实现

Published at: 2015年第1卷第S1期

周支瑞
关键词:

今天奉上临床研究样本量计算系列的第四弹,有关生存资料预后研究的样本量计算----基于Logrank Test的统计效能或样本量估算。生存资料的特点是同时考虑终点事件的出现以及出现终点事件所经历的时间(一般笼统称作生存时间)。对应的统计方法应选择生存分析的方法。由于生存时间一般不呈正态分布,而且需要考虑截尾数据,生存分析有其独特的统计学方法。常用的统计学方法有以下几种。(1)描述性分析:根据样本生存资料估计总体生存率及其他有关指标(如中位生存时间等)。常采用Kaplan-Meier法(乘积极限法)进行分析。对于频数表资料则采用寿命表法进行分析。计算生存率需要考虑时间顺序。(2)比较分析的方法:对不同组生存率进行比较分析,常采用非参数的log-rank检验,检验无效假设使两组或多组总体生存时间分布相同。(3)影响因素分析:通过生存分析模型来探讨影响生存时间的因素,常用的方法为COX比例风险模型。下面我们继续以案例形式进行讲解:

【案例】拟进行某种抗肿瘤新药A对比标准药物B的III期临床试验,根据以往文献报道,已知标准药物B治疗组的总归因病死率是30%。根据II期临床试验的结果新药A治疗组的总归因病死率为15%。本研究计划用2年时间入组患者,计划在入组完成之后继续随访5年,预计试验组与对照组的失访率在5%以内。按照1:1平行设计,取alpha=0.05,power=80%,双侧检验,每组需要多少样本?总计需要多少样本?

第一步,如图依次点击:

图1 依次选择Survival----Logrank----Lakatos---- Logrank Test (Lakatos) [Mortaility]

 

第二步,如图设置参数

图2设置参数

 

参数的具体含义及解释如下:

N(Total Samplesize)表示待计算的样本量,此处为选择项;power=80%表示把握度为0.8,alpha=0.05表示检验水准为0.05;Proportion in Control Group=0.5表示按照1:1分组;M1 (Mortality - Control Group)=30%,Treatment Group Parameter此处计算对照组的病死率/试验组的病死率,前者为0.3,后者为0.15,所以MR(Motality Ratio=M2/M1)此处应填0.5;T0(Survival Time)=5,表示此处为5年归因病死率。Accrual Time=2,表示入组时间为2年;Accrual Pattern=Equal,表示招募患者时单位时间内入组率是相等的;Total Time=7,表示入组时间+随访截止时间共7年;Control Lost&Treatment Lost=0.05,表示试验组与对照组的失访率为5%。

样本量计算结果及报道如下:

Logrank Test Power Analysis

Numeric Results in Terms of Sample Size when the Test is Two-Sided and T0 is 5

 

References

Lakatos, Edward. 1988. 'Sample Sizes Based on the Log-Rank Statistic in Complex Clinical Trials', Biometrics,Volume 44, March, pages 229-241.

Lakatos, Edward. 2002. 'Designing Complex Group Sequential Survival Trials', Statistics in Medicine, Volume21, pages 1969-1989.

Report Definitions

Power is the probability of rejecting a false null hypothesis. Power should be close to one.

N1|N2|N are the sample sizes of the control group, treatment group, and both groups, respectively.

Hazard Ratio (HR) is the treatment group's hazard rate divided by the control group's hazard rate.

Mortality is the proportion dying before time T0.

Accrual Time is the number of time periods (years or months) during which accrual takes place.

Total Time is the total number of time periods in the study. Follow-up time = (Total Time) - (Accrual Time).

Ctrl Loss is the proportion of the control group that is lost (drop out) during a single time period (year or month).

Trt Loss is the proportion of the treatment group that is lost (drop out) during a single time period (year or month).

Ctrl to Trt (drop in) is the proportion of the control group that switch to a group with a hazard rate equal to the treatment group.

Trt to Ctrl (noncompliance) is the proportion of the treatment group that switch to a group with a hazard rate equal to the control group.

Alpha is the probability of rejecting a true null hypothesis. It should be small.

Beta is the probability of accepting a false null hypothesis. It should be small.

Numeric Results in Terms of Events when the Test is Two-Sided and T0 is 5

 

Summary Statements

A two-sided logrank test with an overall sample size of 231 subjects (115 in the control groupand 116 in the treatment group) achieves 80.1% power at a 0.050 significance level to detect a hazard ratio of {NA} when the control group median survival time is a hazard ratio of 00.3000. The study lasts for 7 time periods of which subject accrual (entry) occurs in the first 2 time periods. The accrual pattern across time periods is uniform (all periods equal). The proportion dropping out of the control group is 0.0500. The proportion dropping out of the treatment group is 0.0500. The proportion switching from the control group to another group with a median survival time equal to that of the treatment group is 0.0000. The proportion switching from the treatment group to another group with a median survival time equal to that of the control group is 0.0000.

 

本例共需要231例样本,其中试验组116例与对照组115例。

至此,关于涉生存资料的预后研究的样本量计算演示完毕,

 

笔者|周支瑞 ,复旦大学附属肿瘤医院放射治疗科在读博士。

doi:10.3978/kysj.2014.1.1490

comments powered by Disqus

附件