Assessing predictive probability of success for future clinical trials

Area of Science:

• Pharmacometrics and Biostatistics
• Drug Development and Clinical Trials

Background:

• Data-driven decision-making is vital in drug development.
• Predictive probability of success (PoS) is a key quantitative tool for estimating future trial success.
• Existing PoS methods need enhancement to incorporate surrogate endpoints and multiple historical studies.

Purpose of the Study:

• To investigate and assess a unified Bayesian approach for calculating the predictive probability of success (PoS).
• To extend PoS estimation to include information from surrogate endpoints and multiple historical studies.
• To improve drug development decision-making through integrated PoS assessment.

Main Methods:

• Reviewed PoS based on historical data for the same endpoint.
• Extended PoS to incorporate surrogate endpoint information with a closed-form solution.
• Utilized a Bayesian meta-analytic approach for integrating data from multiple historical studies.

Main Results:

• Developed and illustrated a unified Bayesian approach for PoS estimation.
• Demonstrated the approach's applicability in oncology and immunology trials.
• Provided an R package "PPoS" for practical implementation.

Conclusions:

• The unified Bayesian approach effectively integrates surrogate endpoint data and multiple historical studies for PoS calculation.
• This integrated assessment enhances the reliability of decision-making in drug development.
• The "PPoS" R package facilitates the practical application of this enhanced methodology.