临床试验中自适应设计的urn模型和RPW规则及其模拟

Simulations of urn model and RPW rule of adaptive design in clinical trials

目的:探讨临床试验中设计驱动型自适应设计的优缺点及适用范围.方法:通过对基于概率罐子(urn)模型的随机化胜者优先(RPW)规则的原理分析得出期望分配公式,并据此进行蒙特-卡罗(MonteCarlo)计算机模拟.结果:假定总体成功率分别为pA,pB的两处理临床试验,当pA,pB不相等且pA+pB<1.5时,采用RPW规则能够将更多的受试者分配给成功率较高的处理组.结论:RPW规则可适用于临床试验中有一定响应延迟的情况,通过选择合适的RPW初始参数可减少样本方差,增加试验的稳定性.

AIM： To probe into the advantages, disadvantages and application scope of design-driven adaptive design in clinical trials. METHODS： Expected allocation formula of randomized play-the-winner（RPW） based on urn model in probability was deduced through principle analysis and Monte Carlo simulation was implemented according to it. RESULTS： Suppose the success probabilities of the two treatment groups in clinical trials are PA and PB, PA is not equal to PB and PA ＋P8 〈 1.5, RPW rule can allocate more patients to a better treatment with higher success probability. CONCLUSION： RPW rules can be applied to clinical trials with response delay. The variance can be reduced substantially and the stability of the clinical trial can be enhanced through appropriate selection of starting parameters of RPW.