Abstract:
|
Response-Adaptive (RA) designs, used to allocate patients in clinical trials, have been generalized to the Covariate-Adjusted Response-Adaptive (CARA) design, which balances treatment assignment among a set of covariates while maintaining features of the adaptive design. Challenges may arise in multi-center trials if differential treatment responses and/or effects among sites exist. We first show how allocation probabilities and treatment effectiveness can be adversely affected in balanced and RA designs when multiple sites are not accounted for. We then propose the Site-Adjusted Response-Adaptive (SARA) approach to account for inter-center variability in treatment effectiveness, including both site and site-by-treatment interaction random effects to calculate conditional probabilities. These success probabilities are used to update assignment probabilities for allocating patients between treatment groups as subjects are accrued. Both frequentist and Bayesian models are considered. We compare the balanced and RA cases with our proposed designs and show that the variability in treatment effectiveness is reduced when accounting for clustering during randomization.
|