Statistical Comparison And Research On Simple Randomization, Stratified Blocked Randomization, Minimization And Dynamic Balanced Randomization In Clinical Trial

Clinical trials are systematic researches in human-being (participants or healthy volunteers) to find out the absorption, distribution, metabolism, excretion condition and the adverse effect of the targeted treatment and confirm or explore the effect and safety of the medical treatment. Since the trials'result are closely related to the life of human-being and the invasive impact can be caused by the treatment, any kinds of new drugs or medical devices are seriously checked when reported and reviewed no matter overseas or domestic. These promote the relevant participants including the trial sponsors, principal investigator, clinical experts and statistical experts to ensure the science quality and validity through the entire period of the trial. The random allocation of the trial plays an important part at the beginning of the whole trial and decides whether the trial can be executed in a rational way.There are many methods used in random allocation of clinical trials including simple randomization, permuted block randomization, stratified randomization, dynamic randomization and so on. With the rapid development of new drug inventions domestically and the circumstance that overseas pharmaceutical companies gradually shifted their clinical trial base to China, the form and the content of clinical trials became diversified. Traditional randomization methods cannot meet the existing demand especially in some specific type of trial. Through years of development, statisticians and clinical trials'experts gradually introduced in the dynamic randomization method to the clinical trial and solve the problem that the traditional method cannot solve.This paper aimed to compare this method with simple randomization, stratified blocked randomization and minimization to explore the allocation balance and the randomness, through simulation in samples with prognostic factors distributed in a balanced way and in a random way. And we also explore whether it is rational to use the traditional statistical methods to analyze the data allocated by dynamic balanced randomization by calculating Type I error and power of statistical results.When comparing several allocation methods in balance distributed sample, the simulation results showed that the dynamic balanced randomization method not only can ensure that all groups enrolled the same number of participants, but also can guarantee balanced distribution of prognostic factors among the treatment groups. For the random distributed population, dynamic balanced randomization method can get the comparable results with the minimization in most conditions when comparing balance, and far better than other two methods. When it comes to the small sample size and prognostic factors with multiple levels, our new method showed better results than minimization. As to the randomness of allocation method, the chance of allocation at random in the dynamic balanced randomization is double when comparing to minimization.Through Monte-Carlo simulation in this paper, we can say that the dynamic balanced randomization is a good allocation method in terms of balance, randomness and the applicability of traditional statistical methods, and an ideal randomization method for trials for strict balance of small sample size. Operators of clinical trials should select suitable allocation method based on the characteristics of trials and the characteristics of allocation method.