Adaptive Designs For Dose-finding Studies In Oncology Clinical Trials

Dose-finding study is the early stage of clinical trials,many are phaseâ… or phaseâ…¡, which is designed to know the pharmacokinetics of drug information through small to medium-sized samples of healthy volunteers to determine the maximum tolerated dose of the drug,and then determine the starting dose for the later phase clinical trials.The starting dose will contribute a lot to determine the success of later phase clinical trials.The developments of new drugs are very expensive,and the failure of later phase clinical trials will lead to serious economic losses to the pharmaceutical companies.The subjects of clinical trials are human,so researchers must take into account the ethical requirements,and make scientific analysis and judgments of drug efficacy and safety as quickly as possible with the largest of the speed and grasp, distribute more subjects to the safe and effective dose groups,let subjects benefit to the greatest extent,and this can lower the costs,reduce the time to market.Therefore the use of a better design to early clinical trial is critical.Adaptive designs can achieve the above requirements.Adaptive designs refer to the dynamic modification of certain aspects of experimental design without damage to the validity and integrity of the scientific nature of the design according to trial information that has been accumulated after the start of clinical trials.Adaptive clinical trial designs involves a lot of research,including the adjustment of sample size and the distribution of the proportion of treatment group,increasing the treatment groups,experimental design adjustments,changes of statistical tests, changes in variables and destination of the trial.The traditional design of dose-finding clinical trials is to assign the subjects to dose groups averagely or at a certain percentage of distribution,complete the whole trial in accordance with the pilot sample size while modifications of designs during the trial are not allowed.Therefore,in recent years,there has been a wide range of concerns about adaptive designs from many statisticians and researchers because of their flexible process nature,but it is rarelv applied in real life.Adatptive Design's theory and practice is still in its immature stage.In this study,we explored and discussed some theoretical and practical issues of safety of single-drug,both safety and effectiveness of a single drug and designs for drug combinations.The specific content includes:1.The designs which focus on the safety of single-drug,including parameters and non-parametric methods.Research results showed that:continuous reassessment methods(CRM) can be more accurately estimate the MTD than the traditional method,and also can save a lot of sample size,though it's a little complicated in the operating process,its advantages can not be denied.Because the method can take full advantage of the prior information,updates the model in time with the reaction information from the patients,determine the dose for the next group of patients according to calculations,the clinical course is more time-saving,high efficiency, but this method depends on the selected mathematical model and has a higher requirement for the statisticians.2.During the study of adaptive designs for phaseâ… /â…¡clinical trials which consider a single agent's safety and efficacy simultaneously,we focus on the binary bivariate endpoints.and did many simulations to compare two methods which are all bayesian methods.one is using toxicity and efficacy odds ratios,and the other one is efficacy-toxicity trade-offs.from the simulation result we can gain the conclusion that the first method is more conservative,when most of the doses are over-toxic,it can terminate the trial with a higher probability;Both of the methods act similarly when the probability of toxicity is low;When the probability increases and then decrease while the dose is increasing,the second method acts better,but the second method has a high requirement for the clinicians and statisticians.so we should choose method depend on the condition of the reality.3.In the drug combination dose-finding study,we used Copula regression model, with its two-dimensional data and the fitting between the two drugs combined effects and toxicity of the two drugs combined probability.A computer simulated trial is done to compare Copula Regression Method and the classic CRM method.The results showed that:Copula regression method can estimate the MTD with a higher probability,distribute more patients near MTD,has relatively fewer toxicity events, more secure and conservative,especially when all the dose combinations are over-toxic,Copula regression method terminated the trial earlier..In other words, Copula regression method can be very good to deal with two-dimensional data and the interaction between two drugs,when determine the dose for next group of patients,it can make full use of information we gained during the trial,and treat more patients near the MTD,terminate the trial when it needs.4.The results showed that Bayesian adaptive designs can act well in dose-finding clinical trials;it can not only improve test efficiency,but also can benefit the subjects to the most extent.To sum up,Bayesian adaptive design methods in simulated trials has gotten better results in dose-finding clinical trials.Therefore we hope it can get a wide range of applications.However,it has higher requirements for statisticians and clinicians because of the complicated statistical methods.