Institut für Medizinische Informationsverarbeitung, Biometrie und Epidemiologie-IBE, Ludwig-Maximilians-Universität, München, Germany.
Leukemia. 2011 Sep;25(9):1433-8. doi: 10.1038/leu.2011.116. Epub 2011 May 20.
The aim of this work was to provide guidelines for appropriate statistical analyses regarding most common endpoints in clinical trials on chronic myeloid leukemia: hematologic, cytogenetic and molecular results, failure-free and event-free survival, and progression-free and overall survival. The reasons for the specified recommendations are explained and important issues are outlined by comprehensive examples. Particular attention is paid to the warning of the application of suboptimal methods that may lead to seriously biased results and conclusions. In the presence of a competing risk like death, Kaplan-Meier analysis should not be applied for time-to-remission endpoints. The appropriate method to estimate the probabilities of a time-to-remission endpoint is the calculation of its cumulative incidence function. However, the exact date of remission is hardly known. Detection of remission depends strongly on evaluation frequencies. Complex composite endpoints comprising many events with considerably heterogeneous severity imply difficulties with interpretation. Time-to-remission and complex composite endpoints are not recommended for primary judgment on efficacy. It is rather advisable to investigate remission status at a fixed time point as a primary endpoint, followed by progression-free and overall survival. For patients with the intended remission success at the time point of interest, relapse-free survival provides an additional primary outcome.
本研究旨在为慢性髓性白血病临床试验中最常见的终点（血液学、细胞遗传学和分子学结果、无失败和无事件生存、无进展和总生存）的适当统计分析提供指导。指定建议的原因得到了解释，并通过综合示例概述了重要问题。特别注意警告应用可能导致严重偏倚结果和结论的次优方法。在存在死亡等竞争风险的情况下，不应应用 Kaplan-Meier 分析来评估缓解时间终点。估计缓解时间终点概率的适当方法是计算其累积发生率函数。然而，缓解的确切日期很难得知。缓解的检测在很大程度上取决于评估频率。包含许多严重程度差异较大的事件的复杂复合终点意味着解释困难。缓解时间和复杂的复合终点不推荐作为疗效的主要判断依据。更明智的做法是在固定时间点作为主要终点研究缓解状态，然后研究无进展和总生存。对于在关注时间点有预期缓解成功的患者，无复发生存提供了另一个主要结果。
