Bayesian Clinical Trial Design Course

This course covers background material about Bayesian design and analysis pertaining to randomized clinical trials, and a detailed hypothetical paralle-group two-treatment sequential Bayesian design along with a simulation for learn how the design performs. Bayesian operating characteristics are emphasized, the most important of these being the correctness of the decision regarding treatment effect.

Pre-course reading should take about 4 hours, and the in-class portion will last 2 hours with significant time for questions and discussion. The instructor is Frank Harrell.

Learning Goals

• Become familiar with Bayes’ rule and Bayesian thinking
• Understand the importance of the direction of information flow in conditional probabilities
• Learn many of the advantages of Bayesian clinical trial design and interpretation
• Understand the full power of Bayesian thinking through a detailed clinical trial example with explicit goals factored into Bayesian stopping rules

Pre-course Study

• Introductory Video: Frequentist vs. Bayes
  Note: In addition to saying that the Bayesian approach treats the treatment effect as random, it also allows the true effect to be fixed, were only it possible to know it. An uncertainty distribution for the effect can be thought of as a device that captures how much we know about something we don’t know exactly.
• Chapter 1 through section 1.1.4 of Bayes Rules by AA Johnson, MQ Ott, M Dogucu
• Introduction to Bayesian Statistics by Woody Lewenstein video
• A Bayesian interpretation of a pediatric cardiac arrest trial (THAPCA-OH) by MO Harhay et al
• Figures 1, 2, 3 in Dexamethasone 12 mg versus 6 mg for patients with COVID-19 and severe hypoxaemia: a pre-planned, secondary Bayesian analysis of the COVID STEROID 2 trial by A Granholm et al
• Side-by-side comparison of frequentist and Bayesian approaches and properties
• Flowcharts depicting frequentist and Basesian paradigms
• Football multiplicities

In-Class Topics

• High-level view of Bayes for evaluating treatments
• Background through section 2.3, 2.6, 2.7
• Multiple outcomes and totality of evidence
• Bayesian clinical trial design through 9.5, 9.9, 9.7, 9.8

Bayesian Analysis and Design Resources

• Introduction to Bayes for Evaluating Treatments course notes
• General Resources
• COVID-19 Clinical Trial Design and Analysis and video about sequential designs
• Flowcharts depicting frequentist and Basesian paradigms
• Side-by-side comparison of frequentist and Bayesian approaches and properties
• Simulating Bayesian operating characteristics and how to validate the evidentiary measure at the decision point
• Football multiplicities
• Type 1 probability α vs. probability of a decision error
• Journey from frequentist to Bayesian statistics
• Blog articles
• Bayesian and frequentist reasoning in plain English
• Datamethods topics
• Talks and here
• Stan Forum
• Articles for learning about Bayesian methods
• Bibliography
• Reporting Bayesian results by David Rindskopf
• Improving transparency and replication in Bayesian statistics: The WAMBS checklist
• Bayesian approach to incorporating adult clinical data into pediatric clinical trials by J Ye and J Travis
• Bayesian workflow by Andrew Gelman et al
• Special issue in PharmStats with Editorial provided by Karen Price and Lisa LaVange: Bayesian Methods in Medical Product Development and Regulatory Reviews - Price - 2014 - Pharmaceutical Statistics
• Bayesian Applications in Pharmaceutical Development
• Bayesian and Frequentist Side by Side interactive app by John Kruschke
• Introductory Video: Frequentist vs. Bayes
• Mark Lai’s Course handouts for Bayesian Data Analysis
• Introduction to Bayesian Statistics by Woody Lewenstein video
• Bayesian analytical methods in cardiovascular clinical trials: Why, when, and how by Samuel Heuts et al
• Bayesian reasoning and the prior in court by R Meester and L Stevens