Xie Long-Shen, Lu Hui
Department of Bioinformatics and Biostatistics, School of Life Sciences and Biotechnology, SJTU-Yale Joint Center for Biostatistics and Data Science, Shanghai Jiao Tong University, Shanghai, China.
Front Pharmacol. 2023 Sep 11;14:1186456. doi: 10.3389/fphar.2023.1186456. eCollection 2023.
A delayed treatment effect is a commonly observed phenomenon in tumor immunotherapy clinical trials. It can cause a loss of statistical power and complicate the interpretation of the analytical findings. This phenomenon also poses challenges for interim analysis in the context of phase II/III seamless design or group sequential design. It shows potential to lead researchers to make incorrect go/no-go decisions. Despite its significance, rare research has explored the impact of delayed treatment effects on the decision success rate of the interim analysis and the methods to compensate for this loss. In this study, we propose an analysis procedure based on change points for improving the decision success rate at the interim analysis in the presence of delayed treatment effects. This procedure primarily involves three steps: I. detecting and testing the number and locations of change points; II. estimating treatment efficacy; and III. making go/no-go decisions. Simulation results demonstrate that when there is a delayed treatment effect with a single change point, using the proposed analysis procedure significantly improves the decision success rate while controlling the type I error rate. Moreover, the proposed method exhibits very little disparity compared to the unadjusted method when the proportional hazards assumption holds. Therefore, the proposed analysis procedure provides a feasible approach for decision-making at the interim analysis when delayed treatment effects are present.
延迟治疗效应是肿瘤免疫治疗临床试验中常见的现象。它可能导致统计效能的损失，并使分析结果的解释复杂化。这种现象在II/III期无缝设计或成组序贯设计的中期分析中也带来了挑战。它有可能导致研究人员做出错误的继续/终止决策。尽管其具有重要意义，但很少有研究探讨延迟治疗效应对中期分析决策成功率的影响以及弥补这种损失的方法。在本研究中，我们提出了一种基于变化点的分析程序，以提高存在延迟治疗效应时中期分析的决策成功率。该程序主要包括三个步骤：I. 检测和检验变化点的数量和位置；II. 估计治疗效果；III. 做出继续/终止决策。模拟结果表明，当存在具有单个变化点的延迟治疗效应时，使用所提出的分析程序在控制I型错误率的同时显著提高了决策成功率。此外，当比例风险假设成立时，所提出的方法与未调整的方法相比差异很小。因此，所提出的分析程序为存在延迟治疗效应时的中期分析决策提供了一种可行的方法。
