Tang Zhongwen
a Novartis , East Hanover , New Jersey , USA.
J Biopharm Stat. 2015;25(6):1312-9. doi: 10.1080/10543406.2014.983646. Epub 2014 Nov 7.
An analytical way to compute predictive probability of success (PPOS) together with credible interval at interim analysis (IA) is developed for big clinical trials with time-to-event endpoints. The method takes account of the fixed data up to IA, the amount of uncertainty in future data, and uncertainty about parameters. Predictive power is a special type of PPOS. The result is confirmed by simulation. An optimal design is proposed by finding optimal combination of analysis time and futility cutoff based on some PPOS criteria.
针对具有事件发生时间终点的大型临床试验，开发了一种在期中分析（IA）时计算成功预测概率（PPOS）以及可信区间的分析方法。该方法考虑了截至IA的固定数据、未来数据的不确定性量以及参数的不确定性。预测能力是PPOS的一种特殊类型。通过模拟验证了结果。基于一些PPOS标准，通过找到分析时间和无效性截断值的最佳组合，提出了一种最优设计。
