The Application Of Reinforced Urn Processes In Phase I Clinical Trials And Related Topics

In order to verify the safety and validity of testing drug or treatment in the human being, the whole research of developing the new drug or treatment is composed with the two main parts: the pre-clinical research and the clinical trail. The randomized clinical trial(RCT)becomes a gold standard of medical research during the second half of the 20th century. And the traditional clinical trail also can be divided into the four main phases: phase I–IV trials.The phase I clinical trial is the first of the four phases. The purpose of phase I clinical trial, where a new drug is tested in a small group of patients for the first time, is to find an acceptable dose of toxicity and/or efficacy from the predeterminative dose range to guide the subsequent Phase II clinical trial. The phase I trial should define a standardized treatment schedule to be safely applied to humans and worth being further investigated for efficacy. When the target dose(MTD or MED) has been evaluated, the maximum tolerated dose (MTD) can be defined as the upper border of the safety interval, and the lower border of the therapeutically useful interval can be named after the minimal effective dose (MED).Until now, several practical designs to estimate the target dose of phase I clinical trail have been proposed. The nonparametric approaches, including the UaD (up-and-down) design and dose escalation"3+3"design, have been widely used for simplicity and convenience, though they were not accurate enough to obtain the precise estimation of target dose. However, the relatively precise posterior estimation can be obtained by the parametric methods, such as biased coin up-and-down procedure (BCD), continual reassessment method (CRM), efficient dose escalation with overdose control (EWOC), and etc, but the prior model hypothesis about the unknown dose–response curve and the complicated posterior inference are required.Unlike the above-cited methods, the RUPs approach was introduced to estimate the target dose as a Bayesian nonparametric approach based on generalized Pólya urn. Compared to the parametric methods, not explicit dose-response curve but prior distribution function of target dose has to be required and the precise estimation of target dose can be obtained by RUPs method.In the traditional RUPs design, each subject gets his drugs with the dose levels in a gradually ascending order from the initial dose until he experiences the expected dose response and the stopping rule is satisfied. As soon as the response is detected, no further drugs will be given and the research passes to the next subject. The same procedure will be performed as above. And the study proceeds until each subject has experienced their procedure. In application of the RUPs approach to analyze the whole response records, the posterior point-estimation of target dose can be obtained. As the Bayesian nonparametric approach in the adaptive design, the traditional RUPs design is of the virtues about the simplification and straight-sight in the trail flow. Meanwhile, the precision of the point-estimation is improved in the traditional RUPs design.The previous studies about the RUPs approach focused on the research of its theory and methodology and only precise prior guess was assumed and discussed. In fact, it is not assured whether the prior guess is precise or not before the determination of target dose with application of RUPs approach in phase I clinical trial. Therefore, it needs the further research about how to strike the balance between the ethics and efficiency in application of the RUPs approach in the clinical trail.In this study, we explored and discussed some theoretical and practical issues of the traditional RUPs design by using computer technology and Monte Carlo simulations. The main contents of our study include the modification of the traditional RUPs design process, developing the SAS program for applying the RUPs approach efficiently, and define the workflow and determination rules of obtaining the target dose etc. At the same time, it also has a feasibility study of applying the RUPs approach to analyze the trail records for improving the precision of the target dose. The main works and results of the study are to be introduced as follows.1. The workflow of traditional RUP design is modified. And the modified RUPs design is on the basis of the multiple endpoints RUPs approach.2. Submitting the SAS program in the Monte Carlo simulation,the posterior point-estimation of target dose can be obtained efficiently.3. On the basis of the simulation outcomes, we find that: When the prior guess about the target dose is truly precise, the most precise point-estimation of target dose can be obtained easily. On the other hand, when the prior guess is unclear, the potential value range of the parameters in the prior guess can be shrunk with the information from the subway trail. And the whole fluctuation of the estimation about target dose will remark the optimal pair of the parameters in the prior guess. The least intense fluctuation of posterior estimation can indicate the most precise prior guess efficiently.4. The precise and imprecise prior guesses are both considered and discussed by simulation from the perspective of phase I clinical trial. And the whole workflow of submitting a SAS macro program in application of RUPs method and the determination rule for deciding the target dose on basis of the precise or imprecise guess in phase I clinical trial are also discussed in detail.5. The feasibility study of applying the RUPs approach to analyze the clinical trail records is shown that, in a way, the precision of the target dose will be improved in application of the RUPs approach to re-calculate the estimation of target dose on basis of the response records from the dose escalation trail.The achievements of this study consist in the following four points: Firstly, the process of traditional RUPs design is revised and the modified RUPs design can strike the balance the ethics and efficiency in application of the RUPs approach in the clinical trail. Secondly, the SAS program makes the application of RUPs approach easily. Thirdly, the whole workflow of submitting a SAS macro program in application of RUPs method and the determination rule for deciding the target dose on basis of the precise or imprecise guess in phase I clinical trial are discussed in detail. Finally, through Monte Carlo simulations, it is feasible to improve the precision of the target dose in application of the RUPs approach on basis of the response records from the dose escalation trail in a way.This study mainly discussed several methods and its issues associated with RUPs design on the basis of submitting the SAS program in the Monte Carlo simulation. The application of RUPs approach is also one part of our study. The results of this study can provide reference for further researches of the program of National Natural Science Foundation of China (Adaptive design in clinical trial and its application, No: 30671823).