Proctor Tanja, Schumacher Martin
Institute for Medical Biometry and Statistics, Faculty of Medicine and Medical Center - University of Freiburg, Freiburg, Germany.
Institute of Medical Biometry and Informatics, University of Heidelberg, Heidelberg, Germany.
Pharm Stat. 2016 Jul;15(4):306-14. doi: 10.1002/pst.1758. Epub 2016 Jun 16.
When analysing primary and secondary endpoints in a clinical trial with patients suffering from a chronic disease, statistical models for time-to-event data are commonly used and accepted. This is in contrast to the analysis of data on adverse events where often only a table with observed frequencies and corresponding test statistics is reported. An example is the recently published CLEOPATRA study where a three-drug regimen is compared with a two-drug regimen in patients with HER2-positive first-line metastatic breast cancer. Here, as described earlier, primary and secondary endpoints (progression-free and overall survival) are analysed using time-to-event models, whereas adverse events are summarized in a simple frequency table, although the duration of study treatment differs substantially. In this paper, we demonstrate the application of time-to-event models to first serious adverse events using the data of the CLEOPATRA study. This will cover the broad range between a simple incidence rate approach over survival and competing risks models (with death as a competing event) to multi-state models. We illustrate all approaches by means of graphical displays highlighting the temporal dynamics and compare the obtained results. For the CLEOPATRA study, the resulting hazard ratios are all in the same order of magnitude. But the use of time-to-event models provides valuable and additional information that would potentially be overlooked by only presenting incidence proportions. These models adequately address the temporal dynamics of serious adverse events as well as death of patients. Copyright © 2016 John Wiley & Sons, Ltd.
在对患有慢性病的患者进行的临床试验中分析主要和次要终点时，常用于事件发生时间数据的统计模型已被普遍使用和接受。这与不良事件数据的分析形成对比，在不良事件数据分析中，通常仅报告一张包含观察到的频率和相应检验统计量的表格。一个例子是最近发表的CLEOPATRA研究，该研究在HER2阳性一线转移性乳腺癌患者中比较了三联疗法与二联疗法。在此，如前所述，主要和次要终点（无进展生存期和总生存期）使用事件发生时间模型进行分析，而不良事件则汇总在一个简单的频率表中，尽管研究治疗的持续时间差异很大。在本文中，我们使用CLEOPATRA研究的数据展示事件发生时间模型在首次严重不良事件中的应用。这将涵盖从简单的发病率方法到生存模型和竞争风险模型（将死亡作为竞争事件）再到多状态模型的广泛范围。我们通过突出时间动态的图形展示来说明所有方法，并比较所得结果。对于CLEOPATRA研究，所得的风险比都在相同的数量级。但是使用事件发生时间模型提供了有价值的额外信息，这些信息仅呈现发病率比例可能会被忽略。这些模型充分解决了严重不良事件以及患者死亡的时间动态问题。版权所有© 2016约翰威立父子有限公司。
