Izumi Shoki, Nomura Shogo, Matsuyama Yutaka
Department of Biostatistics, Division of Health Sciences and Nursing, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan.
Section of Biostatistics, Department of Clinical Data Science, Clinical Research & Education Promotion Division, National Center of Neurology and Psychiatry, Tokyo, Japan.
Stat Med. 2025 May;44(10-12):e70112. doi: 10.1002/sim.70112.
In confirmatory randomized controlled trials of patients with metastatic cancer, progression-free survival (PFS) and overall survival (OS) are often used as multiple primary endpoints. The overall hierarchical strategy is a typical multiplicity adjustment method that analyzes PFS once and performs an interim OS analysis at the time of PFS analysis using an alpha-spending function-only if the statistical significance of PFS is demonstrated. A subsequent final OS analysis is conducted if the interim OS analysis does not result in early stopping for efficacy. In this study, we focused on the adjustment of conditional bias (CB) in hazard ratio estimates for OS in both interim and final analyses when a trial applied the overall hierarchical strategy. As CB-adjusting estimators for a single primary endpoint may have limited performance, we extended the conditional mean-adjusted estimator to the case of an overall hierarchical strategy. Motivated by an actual oncology trial, we evaluated the performance of the proposed estimators through a simulation study. In the case of early stopping for efficacy, the CB of the proposed estimator was smaller than that of the existing methods with comparable root mean squared error.
在转移性癌症患者的验证性随机对照试验中，无进展生存期（PFS）和总生存期（OS）常被用作多个主要终点。整体分层策略是一种典型的多重性调整方法，该方法先分析一次PFS，并在PFS分析时使用仅基于α消耗函数进行中期OS分析——前提是PFS具有统计学显著性。如果中期OS分析未导致因疗效而提前终止试验，则随后进行最终OS分析。在本研究中，我们关注的是当一项试验采用整体分层策略时，在中期和最终分析中对OS风险比估计值的条件偏倚（CB）进行调整。由于用于单一主要终点的CB调整估计量的性能可能有限，我们将条件均值调整估计量扩展到了整体分层策略的情况。受一项实际肿瘤学试验的启发，我们通过模拟研究评估了所提出估计量的性能。在因疗效而提前终止试验的情况下，所提出估计量的CB小于具有可比均方根误差的现有方法。
