Earlier this week, we at Cytel enjoyed a riveting in-house discussion on the uses of Bayesian decision rules for Go/No-Go (GNG) decision-making. GNG rules establish the trajectory of a particular clinical program’s development by assessing whether or not a trial has met particular objects (e.g. target regions for PK, PD and safety endpoints.)
Traditionally, statisticians have used p-values and confidence intervals to construct GNG rules. However, moving to the Bayesian paradigm opens up exciting new possibilities for clinical development strategy. Our discussion earlier this week centered around three key benefits of using Bayesian statistics for GNG decision-making:
Firstly, GNG decision-making is most effective when combined with powerful simulations. The Bayesian paradigm seems to provide several benefits for conducting such simulations with accurate results.
Secondly, there is an exciting question arising of how to apply GNG rules to adaptive trials using Bayesian methods. Adaptive trials are often built to be flexible, but this means a greater array of possible trial trajectories. What are key clinical development strategies that benefit from Bayesian methods?
Third, under what conditions have our clients received FDA approval for such GNG rules?
The use of Bayesian methods for GNG decision-rules is in some sense a budding field, but many pharmaceuticals have already met with more than a degree of success in employing them. If you are interested in learning more about the debates within this field, our friends at Merck have put together this Statistical Primer on the use of Bayesian methods for Early Development GNG decisions. They have kindly agreed to allow us to share their primer with you:
If you have questions about using Bayesian methods in establishing your own GNG rules, please don’t hesitate to get in touch.