希望区域法在临床试验适应性设计中的准确性和稳健性研究

The Accuracy and Robustness of the Promising Zone in Adaptive Clinical Trial Designs

目的评估希望区域法应用于适应性设计样本量再估计的准确性及稳健性,为该方法的适用条件提供理论参考。方法以二分类资料为例,以包含期中分析的两阶段适应性试验为框架,采用Monte Carlo方法,在相同的组间响应率率差、同合并率等参数的模拟场景下,比较固定样本量设计、成组序贯设计和希望区域法进行样本量再估计的准确性和稳健性。结果模拟研究显示希望区域法在期中检验效能落入希望区域内进行适应性样本量调整时,Ⅰ类错误与同场景下固定设计和成组序贯设计近似。初始估计样本量被低估时,希望区域法比固定设计的检验效能平均提高约5%,比成组序贯设计提高约8.8%,样本量比固定设计多耗费0.18倍,比成组序贯设计多0.38倍。高估初始估计样本量时,三种设计的检验效能相差仅在1%左右,但希望区域法比固定设计多消耗7.5%、比成组序贯设计多耗费48%的样本量。结论相比于固定样本量及成组序贯设计,希望区域法对于试验的整体检验效能提高不大,其平均样本量均大于相同设定场景下的固定设计及成组序贯设计,使用该设计方法时应谨慎权衡试验收益。

Objectives This study aims to evaluate the accuracy and robustness of the promising zone(PZ)for sample size re-estimation(SSR)in adaptive clinical trials,providing theoretical reference for the applicability conditions of the method.Methods Using binary data as an example,within the framework of a two-stage adaptive trial with interim analysis(IA),Monte Carlo simulation was used to compare the accuracy and robustness of the fixed sample size design(Fixed),the group sequential design(GSD),and PZ for SSR,under the same simulated scenarios of between-group response rate difference and merger rate.Results Simulation studies demonstrated that when the IA result was promising for SSR,the Type I error rate of PZ for SSR was comparable to Fixed and GSD.When the initial estimated sample size was underestimated,the statistical power of SSR was on average approximately 5%higher than that of Fixed designs and approximately 8.8%higher than that of GSD,with a average sample size increase of 0.18 times that of Fixed designs and 0.38 times that of GSD.When the initial estimated sample size was overestimated,the difference in power among the three designs was only about 1%,but SSR consumed 7.5%more samples than Fixed designs and 48%more samples than GSD.Conclusion Compared to Fixed and GSD,PZ is only suitable for scenarios where the initial estimated sample size is underestimated.In scenarios where the initial estimated sample size is overestimated,the PZ does not significantly improve the overall power of the trial.Furthermore,the average sample size of the PZ is higher than that of Fixed and GSD under the same settings.Careful consideration of the trade-off between the benefits and costs of using the PZ in clinical trials.