Bayesian sample size determination for a Phase III clinical trial with diluted treatment effect

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Pharmacoeconomics

Background:

• Phase III clinical trials often face diluted treatment effects compared to Phase II findings.
• Accurate sample size determination is crucial for trial efficiency and reliable results.

Purpose of the Study:

• To propose a Bayesian method for determining Phase III sample size using Phase II data.
• To evaluate the impact of different prior distributions (normal, uniform, truncated normal) on sample size.
• To investigate the 'hook phenomenon' in Bayesian Historical Predictive Power (BHPP).

Main Methods:

• Utilizing normal, uniform, and truncated normal prior distributions for treatment effects.
• Calculating Bayesian sample size based on Bayesian Predictive Power (BPP) and BHPP.
• Conducting numerical simulations to determine sample size and analyze sensitivity.
• Applying the method to the axitinib clinical trial example.

Main Results:

• The Bayesian approach provides a refined sample size by incorporating prior information from Phase II.
• A 'hook phenomenon' was observed for BHPP under specific Phase II sample sizes (n=70) with tested prior distributions.
• Sensitivity analysis revealed the impact of parameter choices on the calculated sample size.

Conclusions:

• The proposed Bayesian sample size calculation method enhances Phase III trial design by leveraging Phase II data.
• Understanding the 'hook phenomenon' is important for accurate power calculations in specific scenarios.
• The methodology offers a flexible framework for sample size determination in clinical trials, demonstrated by the axitinib example.