Institut für Mathematische Stochastik, Georg-August-Universität Göttingen, Goldschmitdstraße 7, 37077, Göttingen, Germany.
Stat Med. 2013 Aug 15;32(18):3055-66. doi: 10.1002/sim.5769. Epub 2013 Mar 18.
The clinical trial design including a test treatment, an active control and a placebo is called the gold standard design. In this paper, we develop a statistical method for planning and evaluating non-inferiority trials with gold standard design for right-censored time-to-event data. We consider both lost to follow-up and administrative censoring. We present a semiparametric approach that only assumes the proportionality of the hazard functions. In particular, we develop an algorithm for calculating the minimal total sample size and its optimal allocation to treatment groups such that a desired power can be attained for a specific parameter constellation under the alternative. For the purpose of sample size calculation, we assume the endpoints to be Weibull distributed. By means of simulations, we investigate the actual type I error rate, power and the accuracy of the calculated sample sizes. Finally, we compare our procedure with a previously proposed procedure assuming exponentially distributed event times. To illustrate our method, we consider a double-blinded, randomized, active and placebo controlled trial in major depression.
临床试验设计包括测试治疗、阳性对照和安慰剂，被称为金标准设计。在本文中，我们为右删失时间事件数据的金标准设计非劣效性试验制定了一种统计方法。我们同时考虑了失访和行政删失。我们提出了一种半参数方法，仅假设风险函数的比例性。特别是，我们开发了一种算法，用于计算最小总样本量及其在治疗组中的最优分配，以便在替代方案下针对特定参数组合达到所需的功效。为了进行样本量计算，我们假设终点呈威布尔分布。通过模拟，我们研究了实际的Ⅰ型错误率、功效和计算出的样本量的准确性。最后，我们将我们的方法与假设事件时间呈指数分布的先前提出的方法进行了比较。为了说明我们的方法，我们考虑了一项在重度抑郁症中的双盲、随机、阳性对照和安慰剂对照试验。
