Interval Estimation in Adaptive Allocations Using Point Estimates

Abstract

Adaptive designs are sometimes used in clihical trials with a goal to allocate a larger number of patients to the better treatment in course of the trial. In the present paper, under the assumption of homogeneous entering patients, an approach of interval estimation to the treatment difference is suggested. The existing approach is the conditional interval estimation which is design independent under the homogeneity of patients. But here we suggest design dependent confidence intervals which is reasonably different under different adaptive designs. For the purpose of illustration we consider three adaptive designs, one is the well known randomized play-the-winner rule, and two new allocation rules. We compare these adaptive designs with a deterministic allocation rule which can be interpreted as the pairwise allocation. Finally we have tried to illustrate the method for all the designs using some real life data set on bladder cancer, which may not have direct relevance with the present set up.