Unit information prior for adaptive information borrowing from multiple historical datasets.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Bayesian Statistics

Background:

• Clinical trials often have multiple historical studies for similar treatments.
• Incorporating historical data improves information gain, efficiency, and treatment evaluation.

Purpose of the Study:

• Propose a novel informative prior, Unit Information Prior (UIP), for adaptive information borrowing from multiple historical datasets.
• Extend the UIP method to handle binary, continuous data, and linear regression models.

Main Methods:

• Develop the UIP based on an information perspective, inspired by Bayesian testing.
• The UIP construction requires only summary statistics, not patient-level data.
• Validate the method through extensive simulation studies and application to real clinical trial data.

Main Results:

• The proposed UIP method demonstrates comparable performance to existing informative priors.
• UIP offers intuitive interpretation and ease of implementation.
• The method successfully adaptively borrows information from multiple historical datasets in a Phase III trial for memantine in Alzheimer's disease.

Conclusions:

• The Unit Information Prior (UIP) provides an effective and practical approach for leveraging historical data in clinical trial analysis.
• UIP facilitates more robust and comprehensive treatment evaluations by adaptively integrating information from past studies.
• The method's reliance on summary statistics makes it widely applicable using commonly reported literature data.