A pragmatic approach for dynamically incorporating predicate device data in prospective diagnostic test studies

Area of Science:

• Medical Diagnostics
• Biostatistics
• Clinical Trial Design

Background:

• Clinical studies are essential for validating new diagnostic tests.
• Historical data from predicate devices can potentially optimize new test evaluation.
• Reducing sample size and trial duration are key objectives in clinical study design.

Purpose of the Study:

• To propose a dynamic Bayesian method for incorporating historical data into new diagnostic test studies.
• To reduce sample size and trial duration by leveraging existing data.
• To enable adaptive trial frameworks for early stopping based on success.

Main Methods:

• Utilizing the Bayesian power prior method with a dynamically calculated power parameter.
• Comparing historical and new data using a one-sided comparison.
• Employing a scaled-Weibull discount function to adjust the effective sample size borrowed.
• Integrating the method within an adaptive trial framework for early success termination.

Main Results:

• The proposed dynamic method allows for effective down-weighting of historical data.
• The approach is pragmatic and conservative, ensuring data integrity.
• The method is demonstrated with an example for detecting Methicillin-resistant Staphylococcus aureus (MRSA).

Conclusions:

• The dynamic Bayesian approach offers a viable strategy to optimize clinical trials for new diagnostic tests.
• Incorporating historical data can lead to more efficient study designs.
• This method supports the development of accurate diagnostic tools with reduced resource utilization.