| Dose-finding is one of the most important activities in the new drug development, which should be encountered at almost every stage, especially in phase I and phase II. The phase I clinical trial is mainly designed for determining the toxicity, which is the MTD; the phase II clinical trial is mainly designed for detection the dose response relationship and the MED. However, the situation in China is that all pay little attention to dose-finding. The dose of the new drug had always been confirmed by the sponsors or related researchers, which was informal. The reasons of the situation can be concluded in two main causes: firstly, the new candidate drug may have both desired and undesired effects, both the efficacy and safety should be explored. All of this means more ethnics challenges will be encountered. Secondly, dose-finding trials always need more sample size and investment than other type clinical trials. So there are little dose-finding related researches, especially in statistical design and analysis of dose-finding. Also dose-finding studies provide challenges for the statisticians. Dose-finding studies belongs to multi-arm trials, which need to control the type I error rate in terms of powerful results. And in adaptive design, the inflation of type I error rate also need to be considered when interim analysis is applied. My research focused on the use of step-wise method based on the phase II dose-finding study, and also discussed some details of its use in the adaptive dose-finding, in order to give some advice to the actual application. The details of my research are as follows:1. When the null hypothesis is true, the student t test, Bonferroni -Holm, Bonferroni-Hochberger, Bonferroni-Hommel, Westfall-Young Bootstrap, Permutational minP and Pre-determined step-down method were compared based on the control of family wise error rate (FWER). On the condition of same sample size, except the t test, other methods can control the FWER effectively. The Pre-determined step-down method is the best one, while others seem more conservative.2. When changed the ratio of allocation of subjects between groups and the numbers of the dose group and the control group, compared the methods mentioned above. The results show whether allocation equally or not, Pre-determined step-down method is more stable than others.3. Based on different dose response models, compared the power on the estimation of MED. The result shows that although Pre-determined step-down method has advantage in getting a more powerful result, it depends on the monotony of the model. Holm method is conservative, but it can not be affected by the type of the model, which means that when the model has an umbrella shape, the method can be used. 4. Improved the adaptive sequence randomization provided by Ivanova. Two different critical bounds were designed for control the number of patient not only in the higher dose groups but also in the lower dose groups. The simulation shows that, when we use the step-wise method in the randomization method it is more conservative in estimating MED.5. Discussed the step-wise method on the problem of two stage seamless design. The first phase provides opportunities for adaptations such as preventing some dose groups into next phase because of futility or efficacy after the interim analysis based on the Pre-determined step-down method. Some other dose groups that haven't got a statistical conclusion will be researched on the second stage .After the second stage study, the P values of the two stages are combined by the fisher combination test.The achievements of this study consist in the following three points: Firstly, we proposed a new strategy on the analysis of the dose-finding study. Secondly, through simulations, we confirmed the conditions of the application of each method, based on the simulation results of the FWER and MED. Finally, we discussed the prospective use of Pre-determined step-down method on the two stage dose-finding study.
|
PDF Full Text Request