The Assessment Of Sample Size Re-estimation Methods In Adaptive Designs

The PhRMA working group defines an adaptive design as a clinical study design that uses accumulating data to decide on how to modify aspects of the study as it continues, without undermining the validity and integrity of the trial. Adaptive design is a dynamic process and allows modifications to the trial in order to improve the success rates and reduce the development costs. Depending upon the types of modification or adaption made, common employed data dependent changes to clinical trials may include, but are not limited to: modifications in inclusion/exclusion criteria, sample size adjustment, changes in hypotheses and/or study endpoints, changes to dosing schedule and duration of treatment, drop the inferior treatment group. In this paper, we focus on the study of methods about adaptive sample size adjustment.Sample size re-estimation methods make it possible to adjust the final sample size according to the information from interim analysis after clinical trials set out. The uncertainty about the parameters at the design stage can be perfected by these methods for the purpose of making the design flexible and minimize the number of patients exposed to inferior or highly toxic treatment groups. In this research, the authors first summarize the techniques of sample size adjustment methods which can be fall into two main groups: internal pilot study and adaptive group sequential design. Simulation studies are then employed to evaluate these methods that mainly include:1.Clinical trials with internal pilot study can be split into two types: blinded sample size re-estimation and unblinded sample size re-estimation. We evaluate the statistical characters of blinded and unblended sample size re-estimation methods under different initial variances, internal pilot sample fractions and statistical analysis procedures.2.Evaluate the statistical characters of CHW method, CRP method and ACR method in adaptive group sequential design, and the application tactics are further discussed here.The main results of this research are as follows:1.If internal pilot study is employed in clinical trials, in the design stage, information about the effect size and variance should be well researched and avoid the situation that the initially estimated variance is much less than the true one; we suggest that the sample size should be adjusted after half of the planned patients have been enrolled; correction t-test is better than classical central t-test and permutation test (especially if the initially estimated variance and/or internal pilot sample size fraction is small). The blinded re-estimation method is preferred and suggested if possible.2.In adaptive group sequential design, the sample size adjustment strategy in CHW method can be described as adjustment based on effect-size ratio and conditional power. The simulation results show that the difference of final sample size and power between two adjustment methods is obvious. In two stage adaptive group sequential design, the results from CHW method, CRP method and ACR method are approximately the same; if ACR method is used, when conditional power is between 0.3 and 0.8 at the first interim analysis, it is not necessary to adjust the final critical stopping boundary, and this strategy can control type I error better.