Novel designs for multi-arm clinical trials with survival outcomes with an application in ovarian cancer
Novel designs for multi-arm clinical trials with survival outcomes with an application in ovarian cancer
复制标题
DOI：
10.1002/sim.1430
复制
发表时间：
2003-07-30
影响因子：
2
通讯作者：
Qian, W
中科院分区：
文献类型：
--
作者：
Royston, P;Parmar, MKB;Qian, W
随着药物开发步伐的加快，几种有前途的治疗方案同时准备在随机III期环境中进行测试并不罕见。各种限制因素，包括将研究结果转移到临床实践所需的时间和狭窄的“机会窗口”，可能使在传统的双臂平行组设计中进行试验以测试这些方案与对照治疗的顺序是不可行的。我们提出了一种试验设计的方法，其基础是在早期阶段消除劣质竞争者，仅允许进入第二阶段的治疗，这些治疗显示出对对照治疗的预定义程度的优势。测试的第一阶段利用已知是有效的中间结果测量或最终结果的替代物的标记物。根据该中间结果测量，将实验组与对照组进行成对比较。在比较中存活的组进入患者累积的第二阶段，最终在主要关注的结果测量上与对照组进行比较。我们展示了如何在实践中实现的设计，考虑假设不同的试验阶段1和2，每个阶段都有自己的操作特点。根据第1阶段和第2阶段的样本量和把握度以及根据双变量正态近似计算的中间和主要结局指标的治疗效应之间的相关性，计算总体操作特征。相关性通过自举先前试验的个体患者数据来估计。我们说明了一般的方法，在一个真实的试验设计的四个新的化疗方案对晚期卵巢癌。中间结局指标是无进展生存期。使用新设计的国际随机对照试验已经在进行中。版权所有（C）2003约翰威利父子有限公司。
With the increasing pace of drug development, it is not unusual for several promising treatment regimens to be ready simultaneously for testing in a randomized phase III setting. Various limiting factors, including the time needed to transfer research results to clinical practice and a narrow 'window of opportunity', may make it unfeasible to perform trials to test such regimens sequentially against a control treatment in a traditional two-arm parallel group design. We present an approach to trial design based on eliminating inferior contenders at an early stage, allowing through to a second stage only treatments that show a predefined degree of advantage against a control treatment. The first stage of testing utilizes a marker known to be a valid intermediate outcome measure or surrogate for the definitive outcome. The experimental arms are compared pairwise with control according to this intermediate outcome measure. Arms that survive the comparison enter a second stage of patient accrual culminating in comparisons against control on the outcome measure of primary interest. We show how the design may be realized in practice by considering hypothetically distinct trials at stages 1 and 2, each with their own operating characteristics. The overall operating characteristics are computed from the stage 1 and 2 size and power and the correlation between the treatment effects on the intermediate and primary outcome measures according to a bivariate Normal approximation. The correlation is estimated by bootstrapping individual patient data from previous trials. We illustrate the general approach in a design of a real trial of four new chemotherapy regimens for advanced ovarian cancer. The intermediate outcome measure is progression-free survival. An international randomized controlled trial using the new design is already under way. Copyright (C) 2003 John Wiley Sons, Ltd.