Abstract:

AIM: In clinical trials, external clinical trial results sometimes trigger unplanned interim decisions, which typically involve early termination and sample size re-estimation. This article aims to evaluate the statistical performance of unplanned sample size re-estimation methods in fixed-sample clinical trials. METHODS: Based on the introduction of unplanned sample size re-estimation methods, including the fixed PP method and the Reoptimization method, a Monte-Carlo simulation study was conducted by setting different true treatment effect between groups, and comparing them with the traditional CPZ method. Statistical performance was evaluated using the Type I error rate, prediction accuracy, expected power, expected sample size, and average sample size. RESULTS: The Type I error rates of the three sample size re-estimation methods were well maintained at the preset level, but only the Reoptimization method maintained the expected power. Furthermore, when the initial estimated treatment effect was equal to or lower than the true treatment effect, the sample size estimated by the Reoptimization method was larger than that of the fixed PP method. When the initial estimated treatment effect was higher than the true treatment effect, the sample size estimated by the Reoptimization method was smaller. CONCLUSION: The Reoptimization method demonstrates better statistical performance and is recommended for unplanned sample size re-estimation in fixed-sample design clinical trials.

Key words: fixed-sample design, unplanned sample size re-estimation, conditional power, conditional error principle