Exploration On Several Aspects Of Adaptive Design In Clinical Trial

The subjects of clinical trial, thoroughly different from that of traditional medical experiment, are human beings. So it is much more essential for researchers and clinicians to make correct decisions about the efficacy and safety of the new drug as soon as possible in conducting clinical trials. How to develope and refine methodological methods for clinical trial with the intent of permitting mid-trial design modifications which are based on interim information from inside or outside the trial to lower the costs and reduce the time to market. Adaptive design, much different from conventinal design types for clinical trias, has been proposed just for this kind of purpose.The traditional randomized clinical trial(RCT)have attracted many statisticians'interest for a long time because of it's simplicity and efficiency. But it does not offer the flexibility to make data-driven adaptive changes to the trial design during the course of the trial. However, drug development is a sequence of complicated decision-making processes, there are many options at each stage and decisions are partially dependent on the prior information, so it should be essential for the design of clinical trial to be flexible and interchangeable. Adaptive design allows modifications to some aspects of the trial after its initiation without undermining the validity and integrity of the trial. The modifications to a trial of adaptive design include a host of aspects, such as sample size re-estimation, early stopping due to efficacy or futility, adaptive randomization, dropping inferior treatment groups, altering study objectives (for example, switching from superiority to non-inferiority or equivalence, or vice versa), redefining the primary endpoint, inserting or skipping interim analysis, etc.Recently, many papers and discussions about adaptive design have sprung in journals showing a vast interest in this topic of researchers who work in the field of clinical trial. However, there are many problems in the design and in the corresponding statistical analysis that hold back the application of adaptive design in practice. In this study, we explored and discussed some theoretical and practical issues of response-adaptive randomization and sample size re-estimation in adaptive desin by Monte Carlo simulation experiments and with some practical data analysis of clinical trials. The main works and results of the study are to be introduced as follows.1. The influence of three parameters of RPW(u,a,b) rule on the treatment allocation proportion was assessed by means of Monte Carlo simulation. The results indicated that the difference of allocation proportions between two treatment groups decreases with the increase of u value and trend towards zero when u become very large. However paramenter a has an inverse influence of that of u, i.e. the difference of allocation proportions between two treatment groups increases with the increase of a value. Compared with parameter a and u, the influence of parameter b is more complex, the difference of allocation proportions between two treatment groups increases as b increases, but the extent of this influence is too low to be concerned. In addition, the treatment allocation proportions were investigated through computer simulation experiment under different situations of treatment successful probabilities and the randomization results of parameter's method and non-parameter's methods were compared.2. Issues of ethics and statistics concerning adaptive design in clinical trials were discussed. For decades, the ethical tension in clinical trial design has been characterized as that between individual and collective ones. But how to balance them is still an open issue. We compared the strengthes of two methods, i.e. blocked response adaptive randomization (BRAR) and altering the u value of the RPW rule (RAR), in mitigating the conflict between collective and individual ethics, by means of Monte Carlo simulation. Conclusions can be drawn from the results that there is no significant difference between the strengthes in balancing the two aspects of ethics and between the influences on accrual bias of these two methods. Because BRAR maybe more complicated than RAR to carry out in practice and may trouble the statistical analysis of the trial, we preferred RAR to BRAR.3. With the practical data of a clinical trial of new drug treating oral ulcer, changes on individual ethics and the influence of response-adaptive randomization on the power of statistical test and several prognostic variables were explored trough computer simulation experiments. It was showed in the results that response-adaptive randomization could be utilized to increase the probability of a patient to receive what would eventually be shown be a superior treatment, which promote the individual ethics in clinical trial. Results also showed that between-group balances of possible prognostic variables can be easily achieved with response-adaptive randomization.4. Given the study time were from 6 to 12 monthes and the interview time were from 7 to 56 days, we calculated the allocation proportions between the two treatments groups through simulation experiments. The outcomes indicated that the effects of study duration and interview duration on randomization results exactly exist, but they were not as serious as we assumed for the differences of effective modulus were not significant. It can be concluded that the Friedman's urn model would be feasible for many clinical trials with tudy duration and interview duration in the given situations.5. Based on the theorem of sample size re-estimation and with the practical data of the phaseâ…¡andâ…¢clinical trials of a new drug, we calculated the conditional power of statistical test with the actual sample size of the phaseâ…¢clinical trial and get a reasonable explaination for the unexpected results of this project. Addationaly, we simulated the distributions of the re-estimated sample sizes by bootstrap method and the results showed that the distributions of re-estimated sample sizes were much skewed distributions with very large variations. When the original sample size was small, these trends will be more obvious and the reliability of the re-estimated sample sizes was lower. Based on the results of simulation experiments in this study, we gave some suggestions for using sample size re-estimation method in clinical trials.The achievements of this study consist in the following three points: First, we disscussed the conflict between collective and individual ethics by Monte Carlo simulation and compared the effects of the two methods to migitate the ethical tension in clinical trial. Second, with the practical data of a clinical trial of a new drug, we explored the use of response-adaptive randomization. Simultaneously, we observed the influences of study duration and interview duration on response-adaptive randomization results and identified the numerial relationship among them. Third, we simulated the distributions of the re-estimated sample sizes by bootstrap method and obtained results and conclusions that were meaningful to some extent for the design of clinical trials.This study mainly discussed several issues about response-adaptive randomization and sample size re-estiomation, in the common frame of so called'adaptive design', and introduced some other related concepts. This study can provide some references for the application of adaptive design in practical clinical trials. Furthermore, the results of this study can be helpful for further researches of the program of National Natural Science Foundation of China (Adaptive design in clinical trial and its application, No: 30671823).