| Objective:To compare the performance of traditional design and bayesian adaptive design in maximum tolerated dose exploration under different experimental conditions,so as to provide scientific basis and application guidance for rational selection of dose exploration methods.Methods:(1)Literature research method:Using databases such as CNKI database,Wanfang and Pubmed to consult relevant data such as maximum tolerated dose exploration methods and so on.To collect and collate the data related to the maximum tolerated dose exploration test and provide scenarios for the simulation and case study.(2)Experts meeting method:Experts in this field are invited to discuss the clinical requirements of this study and suggestions for improving the application of methods.(3)Statistical analysis:The maximum tolerated dose exploration methods use of bayesian statistics and use the rank-sum ratio method to comprehensively evaluate the test performance of the dose exploration methods.Data processing and statistical analysis are mainly completed in R software and SAS 9.4 software,and Graphipad Prism 8.3 drawing software is used for chart drawing.(4)Simulation method:Simulation is used to compare the experimental performance of the dose exploration methods by assuming the known dose exploration test data and creating test scenarios similar to the real clinical trials.(5)Empirical research method:To verify the results of comparison between theory and simulation by using actual clinical data and combining with experimental operation process.Results:(1)Theoretical study on maximum tolerated dose exploration methods:The principles and characteristics of various methods are clarified and compared theoretically.(2)Simulation comparison of maximum tolerated dose design methods.Numerical simulation method is used to compare the performance of MTD identification correctness,safety and test efficiency under different sample size,cohort sizes,number of dose levels,target toxicity probability and dose toxicity scenarios.The traditional design represented by 3+3 performs poorly.In the bayesian adaptive design,the model design represented by CRM performs well.BOIN,mTPI,and Keyboard are three kinds of model-assisted designs.The above three kinds of experimental performances are between traditional 3+3 and CRM.Among them,mTPI performs slightly worse than BOIN and Keyboard.Different experimental conditions will have different effects on the three experimental performances of different dose exploration methods.(3)Case studies of the maximum tolerated dose design methods:Using actual clinical data of drugs to verify the results of simulation comparison and compare the ease of operation,accuracy of MTD estimation and flexibility of different methods.The accuracy,safety and test efficiency of MTD estimation are consistent with the simulation results.The traditional design represented by 3+3 has the best performance in ease of operation.CRM performs well in MTD estimation accuracy and flexibility,but is the least easy to operate.Model-assisted designs,such as BOIN,are easy to operate and have flexibility between 3+3 and CRM.(4)According to the comparative results of this study,the rank-sum ratio method is used to evaluate the dose exploration methods.According to the purpose of the experiment,the advantages and disadvantages of each method and the applicable conditions,the selection and application to form the suggestions of dose exploration methods.Conclusion:In this study,simulation and empirical study are combined to compare the experimental performance of the dose exploration methods in multiple dimensions,considering the experimental conditions and other influencing factors.The traditional design represented by 3+3 has obvious limitations on the correctness of MTD recognition in addition to the outstanding advantage of easy operation.The bayesian adaptive design method has an excellent performance in the correctness of MTD recognition.Among them,the model design represented by CRM has a complex operation process,but other experiments have a good performance.Model-assisted designs overcome the obvious shortcomings of the traditional design and model design,and the experimental performances are in between.So it can be regarded as a good design method without considering CRM.Combined with the comparison results,this study summarize and put forward suggestions on the selection and application of dose exploration methods under different conditions.It can provide feasibility reference for MTD exploration experiment. |
PDF Full Text Request