Bayesian Implementations in Drug Development: Why Progress Is Slow

• Activity Number: 228
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Beat Neuenschwander*+
• Companies: Novartis Pharma
• Keywords: Bayesian ; frequentist ; likelihood ; statistical engineering ; statistical science ; statistical thinking

Abstract

Bayesian Implementations in Drug Development: Why Progress is Slow

In many areas of the empirical sciences the Bayesian approach is on the rise. This trend has also caught up with drug development. However, statisticians involved in clinical trials face difficulties when trying to implement the Bayesian approach.

The difficulties have various causes, related to statistical thinking, science, and engineering. First, the Bayesian approach is ambitious due to the fact that it extends the conventional paradigm by adding prior and predictive thinking. These extra dimensions of statistical thinking are important components of Bayesian reasoning, but they are also challenging. Second, statistical science disagrees over fundamental metrics of measuring evidence (frequentist, likelihood, Bayesian). A broader view that requires good properties for various key metrics (instead of optimal properties of a single metric) would seem a way forward. And third, statistical engineering requires a considerable amount of resources, which is often underestimated in practice.

This talk will provide examples on how to address these issues and put Bayesian reasoning into practice.