Bayesian clinical trial design using historical data that inform the treatment effect

Area of Science:

• Biostatistics
• Clinical Trial Design
• Bayesian Inference

Background:

• Clinical trials often face challenges in sample size determination, especially when historical data are available.
• Integrating prior information from historical data can improve the efficiency and power of new clinical trials.
• Traditional frequentist methods for sample size calculation may not fully leverage available historical data.

Purpose of the Study:

• To develop a Bayesian methodology for sample size determination in clinical trials using historical data.
• To calibrate prior informativeness and ensure appropriate statistical power and type I error control.
• To provide a framework for designing efficient clinical trials by incorporating historical treatment effect information.

Main Methods:

• A simulation-based framework was developed to calibrate prior distributions and determine optimal sample sizes.
• Bayesian generalizations of type I error control and power were defined using null and alternative sampling priors.
• A partial-borrowing power prior was used to construct a Bayesian hypothesis test and summarize information borrowing.

Main Results:

• The proposed methodology effectively determines sample size, balancing statistical power and type I error control.
• Simulation studies showed the partial-borrowing power prior approach is as efficient as complex meta-analytic priors.
• The methodology was successfully applied to design a melanoma clinical trial with a time-to-event endpoint.

Conclusions:

• Bayesian sample size determination using historical data offers a robust and efficient alternative to traditional methods.
• The developed methodology provides a practical framework for designing clinical trials with improved statistical properties.
• This approach facilitates optimal resource allocation and enhances the likelihood of detecting true treatment effects.