A New Family of Covariate-Adjusted Response-Adaptive Randomization Procedures for Precision Medicine
Open Access DepositedIn most clinical trials, patients accrue sequentially and need to be assigned to different treatment groups. Previous studies of randomization procedures include complete randomization, restricted randomization, response-adaptive randomization (RAR), covariate-adaptive randomization (CAR) and covariate-adjusted response-adaptive randomization (CARA). With the development of precision medicine, information about biomarkers is usually available and should be included in the randomization procedure. In statistical analysis, biomarkers are mathematically treated as covariates and we classify biomarkers into predictive and prognostic covariates according to their roles: predictive covariates are used to select the suitable treatments and prognostic covariates should be balanced across treatment assignments. Under this setting, we find the drawbacks of existing designs and propose a new family of CARA design for precision medicine, considering both predictive covariates and prognostic covariates. The new family of CARA design integrates both covariates balance and target allocation that can not only assign more patients to better treatments based on the predictive covariates, but also balance prognostic covariates. More specifically, a new Weighted Balance Ratio (WBR) for prognostic covariates is defined within each strata of predictive covariates and is incorporated into Doubly-Adaptive Biased Coin Design (DBCD) method with Hu and Zhang’s allocation function. In Chapter 2, we propose a simplified version of the new family of CARA design: only consider prognostic covariates Z without predictive covariates X and look into this design in another way: to balance prognostic covariates in response-adaptive designs. The simplified CARA design integrates both prognostic covariates balance and an unknown target allocation proportion. That means, we construct response-adaptive designs while balancing covariates at the same time. Simulation results and theoretical results regarding the simplified design is studied in Chapter 2. A redesign of HIV transmission trial is also shown in Chapter 2. Under the setting of precision medicine, we propose the general new family of CARA design in Chapter 3, considering both prognostic covariates Z and predictive covariates X. The new family of CARA randomization procedure can (i) assign more patients to better treatment groups by calculating the optimal treatment allocations within stratum of their predictive covariates X and (ii) provide a more balanced treatment assignments with regarding to the prognostic covariates Z. Simulation results and theoretical results regarding the general design are studied in Chapter 3. A redesign of the Stroke Prevention in Atrial Fibrillation (SPAF) Trials is also shown in Chapter 3. In Chapter 4, we further study the proposed Weighted Balance Ratio (WBR) and discuss the possible incorporation of WBR into difference randomization. A comparable study is shown in Chapter 4.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.