In the Interim... Bayesian Statistics in Clinical trials: The Past, Present, and Future
9 snips
Nov 24, 2025 Don Berry, a pioneering Bayesian statistician with over 50 years of experience, joins his family members Nick and Lindsay Berry to discuss the fascinating evolution of Bayesian methods in clinical trials. They explore the ethical challenges of trial design, comparing patient treatment and learning. Lindsay shares insights from implementing adaptive trials during the COVID-19 pandemic, while Nick highlights the computational advancements of MCMC that made Bayesian modeling practical. Together, they delve into the future of Bayesian statistics, touching on its implications for AI and individualized medicine.
AI Snips
Chapters
Books
Transcript
Episode notes
Treating Patients While Learning
- Clinical trials balance treating current patients and learning for future patients, like a sequential bandit problem.
- Don Berry argues we can and should combine treatment and learning rather than treating trials as only for learning.
Rewriting I-SPY 2 Code Overnight
- Scott describes rewriting lost I-SPY 2 trial code over a long weekend to recover Bayesian trial logic.
- He emphasizes building Bayesian trial algorithms is hard but becomes faster with experience.
MCMC Made Bayes Doable
- The advent of MCMC and increased computing power made Bayesian methods practically usable for complex models.
- Scott Berry says MCMC transformed Bayesian methods from philosophical to computationally feasible in trials.
