.png){kind=icon}
.png){kind=icon}
Return
Prospectively Specified Adaptive Bayesian Borrowing: Considerations, Methodologies, and Implementations
.png){kind=icon}
PRE
AI
DOI：10.1002/pst.70051.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In clinical research, it is increasingly difficult to conduct fully powered and well-balanced randomized controlled trials, particularly when studying rare or devastating diseases and pediatric patients. While Bayesian methodologies are very useful for leveraging historical control data to meet some of these challenges, many practical and statistical concerns emerge when prospectively specifying a design to implement Bayesian methods. In this article, we discuss these concerns and propose novel methods to ensure statistical rigor when applying Bayesian methodology. A novel adaptive Bayesian borrowing (ABB) method proposed here borrows from historical control data to increase the precision of the control arm based on the observed congruence of the historical and current data. The method would also enable an adaptive increase of sample size to accommodate accumulating information. We demonstrate that this approach can be prospectively specified and provides a statistically rigorous and transparent inference while mitigating the risk of potential conflict between historical and current control data, as well as misspecifications of variability in the endpoints.
Keywords:
adaptive Bayesian borrowing
blinded sample size re-estimation
clinical trials
leveraging historical control
MAP priors