Bayesian semiparametric meta-analytic-predictive prior for historical control borrowing in clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Bayesian Inference

Background:

• Clinical trial efficiency can be improved by incorporating historical control data.
• Existing Bayesian methods for historical control borrowing require pre-specification of prior parameters.
• This can be problematic due to potential heterogeneity or conflict between historical and current trial data.

Purpose of the Study:

• To propose a novel Bayesian semiparametric meta-analytic-predictive prior for flexible historical control borrowing.
• To develop a method for estimating the prior effective sample size (ESS) for quantifying borrowed information.
• To demonstrate the effectiveness of the proposed method in improving clinical trial design and analysis.

Main Methods:

• Developed a Bayesian semiparametric meta-analytic-predictive prior using a Dirichlet process mixture prior.
• Generalized a method for estimating the prior effective sample size (ESS).
• Conducted extensive simulation studies and analyzed a phase II proof-of-concept trial in ankylosing spondylitis.

Main Results:

• The proposed method adaptively learns the relationship between historical and current control data, relaxing parametric assumptions.
• The estimated ESS provides an intuitive quantification of borrowed information, aiding in prior tuning.
• The new methodology demonstrated superior performance compared to existing methods in simulations and a real-world trial.

Conclusions:

• The robustified meta-analytic-predictive prior alleviates the need for pre-specifying the amount of borrowing.
• This provides a more flexible and robust approach for integrating historical data from multiple sources in clinical trials.
• The method enhances the efficiency and power of clinical trial designs by optimizing the use of historical control information.