Comparison Of Several Testing Strategies For All-randomized Design In Targeted Clinical Trials

Objective: As the development of molecular biology and gene technology,the research and development of targeted drugs is increasing attention,and the design of clinical trials based on targeted drugs is constantly proposed.The purpose of this article is to evaluate statistical performance of four testing strategies for allrandomized design in targeted clinical trials and to estimate the sample size of different strategies in order to provide basis for the selection of analysis methods in targeted drugs research.Methods: Simulating the data based on continuous and dichotomous outcome via Monte Carlo Method and considering the influence of biomarker positive rate and the efficacy of targeted drugs in different target subgroup.Then comparing power and type I error rate and estimating the sample size of sequential subgroup-specific strategy,sequential biomarker-positive and overall population strategy,marker sequential test design and fall-back design.Results: When the biomarker-positive prevalence is high,there is little difference in power of the four strategies for the positive subgroup analysis.While in the overall population analysis,sequential subgroup-specific strategy and MaST have higher power.When the biomarker-positive prevalence is low,if the efficacy of targeted drugs in different marker-status populations is quite different,in the positive subgroup analysis,sequential subgroup-specific strategy,sequential biomarkerpositive and overall population strategy and MaST have higher power,and MaST has highest power in the overall.If the targeted drugs in the different markers of the state of the crowd less difference,fall-back design and MaST have higher power in the positive subgroup analysis and power of sequential subgroup-specific strategy and MaST is higher in the overall analysis.Sequential subgroup-specific strategy and sequential biomarker-positive and overall population strategy can control type I error in the positive subgroup analysis,while sequential subgroup-specific strategy and MaST can in the overall.In sample size estimation,sequential subgroup-specific strategy and sequential biomarker-positive and overall population strategy is only affected by the biomarker positive rate and have no association with different target population curative effect.The sample size of MaST and fall-back design increased with the increase of efficacy discrepancy,while decrease with increased biomarker positive rate.Conclusion: In the all-randomized clinical trials of targeted drugs,exploring the efficacy of targeted drugs in the positive population,we recommend sequential subgroup-specific strategy and sequential biomarker-positive and overall population strategy,and fall-back design is the optimal approach only if there is a clear difference between the efficacy of the targeted drugs in different marker populations and the low positive rate of the population marker.While MaST is the first recommended design in the overall population study.In targeted positive people efficacy study,when targeted drugs curative effect difference in positive and negative subgroup and biomarker positive rate is low,MaST and fall-back design can significantly reduce the sample size.