Biostatistics Case Study #01: Interim Analysis using Adaptive Randomization

Adaptive randomization refers to any scheme in which the probability of treatment assignment changes according to assigned treatments of patients already in the trial. The covariate adaptive randomization (CAR) is usually used instead of pure randomization to reduce the covariate imbalance between treatment groups in clinical trials. Allocation probability for the covariate adaptive randomization is adapted overtime during the trial based on the cumulative information about baseline covariates and treatment assignments

Covariate Adaptive Randomization

A phase II, prospective, multi-center, randomized, 4-week, double-blind, placebo-controlled, designed to determine the safety, tolerability of fixed oral dose of Study treatment in patients with chronic schizophrenia. To implement Covariate Adaptive Randomization based on the interim analysis to increase the probability of success.

Methods addressed by biostatisticians to reduce the covariate imbalance in experiments is Frane Method

The data available is the covariate information of 60 subject. Considering the covariate to be Gender and Country, data will be summarized as below for two treatment Groups. Temporarily assign the new patient to treatment group A first. Calculate the Pearson’s χ2 Goodness-of-Fit test statistic for the covariate groups to which the new patient would belong. Identify the maximum χ2 test statistic among all the covariate groups. Now temporarily assign the new patient to Placebo group and repeat above steps. Identify the minimum X2 test statistic over all the identified test statistics and assign the new subject to that corresponding.

New patient allocated

Using SAS code and the prior information available on covariates of the trial, we assigned the patient to ‘Treatment A’ or ‘Placebo’ based on the nature of covariate of the new patient.


To know more, write to Click here to download this case study.

Submit your response

Your email address will not be published.