微信直播

AME循证杂谈020|两样本均数比较的样本量计算——基于PASS软件实现

Published at: 2015年第1卷第S1期

周支瑞
关键词:

循证杂谈今日重出江湖,第一弹,如题!

 

样本量计算在临床研究设计阶段占有重要位置,如果读者熟悉随机对照临床试验的报告规范----CONSORT声明,其中一条款要求临床研究的报告应该报告样本量估算的过程。如何正确估算样本量?是否有合适的工具可以帮助我们估算样本量?本文将介绍一种权威、好用的样本量估算软件,并以案例演示的形式向大家展示两样本均数比较的样本量计算过程。

【案例】一项临床随机对照试验研究新型降压药物A治疗高血压的作用,对照药物选择临床确定有效的盐酸贝那普利片(洛丁新)。如果A药降低收缩压的能力低于洛丁新10mmHg,则认为该药无临床应用价值,若不低于10mmHg则认为该药有临床应用价值。估计新药A实际降低收缩压能力最多低于洛丁新5mmHg。根据既往文献报道洛丁新组降低收缩压的标准差为18mmHg,根据预实验结果,新药A降低收缩压的标准差为18mmHg。按照1:1平行对照等效设计,选用alpha=0.05,power=90%,双侧检验,每组需要多少样本?总计需要多少样本?

题外话,当我还是一只菜鸟的时候,有人告诉我用统计课本上的公式和附表去计算,偶尔还需要使用科学计算器,当时信以为真,并真的实践了。但那是一个非常痛苦的经历,尤其是像我这种丢了数学很多年的医生。工欲善其事,必先利其器!隆重介绍本文的主角----PASS(power analysis and sample size)软件,这绝对是一款计算样本量的必备神器。下面我们就把上文的案例计算一遍,计算过程如下:

第一步,打开软件,界面如下

图1 软件界面

 

第二步,依次做如图所示选择:

图2 依次点开Means----Non-Inferiority---- Non-Inferiority Tests for Two Means [Differences]

 

第三步,如图填入参数

图3 依次填入参数,正确填入后点击坐上角Run

 

参数的具体含义及解释如下:

N1表示待计算的试验组样本量,此处为选择项;power=90%表示把握度为0.9,alpha=0.05表示检验水准为0.05;Use R,R=1.0的含义是:试验组与对照组按照1:1分组;NIM (Non-Inferiority Margin)=10mmHg,本例中只要试验组降低血压值不低于10mmHg,我们就认为试验组药物有临床应用价值,这就是所谓的非劣效的限值;D (True Difference)= 5mmHg,本例中估计新药A实际降低收缩压能力最多低于洛丁新5mmHg,所以此处实际均数差填5 mmHg;S1 (Standard Deviation Group 1)或者 S2(Standard Deviation Group 2)表示试验组及对照组的标准差,此处根据文献报道及预实验已知试验组与对照组的标准差均是18 mmHg;Higher Means Are=Worse,此处应该做出正确选择,方向不能反了,本例中如果血压均值越高表示降压药物效果较差。

 

样本量计算完整结果如下:

Power Analysis of a Non-Inferiority Test of The Difference of Two Means

Numeric Results for Non-Inferiority Test (H0: Diff >= NIM; H1: Diff < NIM)

Higher Means are Worse

 

Test Statistic: T-Test

References

Chow, S.C.; Shao, J.; Wang, H. 2003. Sample Size Calculations in Clinical Research. Marcel Dekker. New York. Julious, Steven A. 2004. 'Tutorial in Biostatistics. Sample sizes for clinical trials with Normal data.' Statistics in Medicine, 23:1921-1986.

 

Report Definitions

Group 1 is the treatment group. Group 2 is the reference or standard group.

 

Power is the probability of rejecting a false null hypothesis.

 

N1 is the number of subjects in the first (treatment) group.

 

N2 is the number of subjects in the second (reference) group.

 

NIM is the magnitude of the margin of non-inferiority. Since higher means are worse, this value is positive and is the distance above the reference mean that is still considered non-inferior.

 

D is the mean difference at which the power is computed. D = Mean1 - Mean2, or Treatment Mean – Reference Mean.

 

Alpha is the probability of a false-positive result.

 

Beta is the probability of a false-negative result.

 

SD1 and SD2 are the standard deviations of groups 1 and 2, respectively.

 

Summary Statements

 

Group sample sizes of 223 and 223 achieve 90% power to detect non-inferiority using a one-sided, two-sample t-test. The margin of non-inferiority is 10.000. The true difference between the means is assumed to be 5.000. The significance level (alpha) of the test is 0.05000. The data are drawn from populations with standard deviations of 18.000 and 18.000.

 

共需要446例样本,试验组与对照组各223例。最后一段报告稍作修改放入论文中即可,大功告成!

 

笔者|周支瑞 ,复旦大学附属肿瘤医院放射治疗科在读博士

 

下期预告

样本量计算第二弹预告

两样本率比较的样本量计算----基于PASS软件实现

 

走过北上广,AME College 精彩仍在继续。下一批城市站点现已敲定——青岛、南京、昆明,哪一个是你所在的城市?名额有限,赶紧扫描下方二维码或点击“阅读原文”报名参加吧!

 

Dec.
19-20

青岛站
地点:青岛大学附属医院

Dec.
26-27

南京站
地点:江苏省人民医院

Jan.
9-10

昆明站
地点:云南省肿瘤医院

doi:10.3978/kysj.2014.1.1427

comments powered by Disqus

附件