1. DIA BSWG
© 2015 DIA, Inc. All rights reserved.

2. Outline Brief Overview of DIA BSWG Bayesian Survey Results Overview
Areas of Focus
© 2015 DIA, Inc. All rights reserved.

3. Value of Bayesian Approach
Emulates how we naturally think
Addresses questions directly
Naturally aligned with adaptive designs
Creates environment to evaluate how others are using prior knowledge
Yields more appropriate sample sizes (can result in saving time and money)
Provides more integrated clinical plan
Increases understanding of uncertainty to enable more informed decisions throughout medical product development
Ability to incorporate prior information from different sources
Natural for evidence synthesis or meta-analysis
Handling multiplicity through borrowing strength and hierarchical modeling
Appealing in dealing with rare events as the model modulates the extremes
Ability to handle complex problems via unified modeling, taking all the uncertainty into appropriate account
Allowing direct probability inferences on different scales
Rev. Thomas Bayes 3

4. Some Key Hurdles
Increased acceptance could transform the future of drug development, but there are some key barriers that currently result in limited uptake, including:
Perceived lack of acceptance from regulatory, including lack of acceptance of simulations to demonstrate control of type 1 error and lack of acceptance for formal use of prior knowledge;
Lack of acceptance for use of Bayesian methods in confirmatory study as basis of approval;
No clear path for timely communication with regulators regarding technical aspects associated with adaptive design and/or Bayesian approaches;
Limited expertise regarding adaptive design and Bayesian methods (computation, methods, knowledge of designs, etc.); and
Internal resistance to change.

5. DIA BSWG: Who are we?
Group of representatives from Regulatory, Academia, and Industry, engaging in scientific discussion/collaboration
facilitate appropriate use of the Bayesian approach
contribute to progress of Bayesian methodology throughout medical product development

6. Opportunity Statement
Bayesian methods provide framework to leverage prior information and data from diverse sources.
Bringing together academic, industrial, and regulatory representatives is essential to overcome hurdles.
Improved patient outcomes
Provides opportunity to influence
proactively by engaging in
scientific discussion

7. Subteams (chairs) Safety (Karen Price/Amy Xia)
Use of historical data/prior specification (John Zhong/Satrajit Roychoudhury)
Non-inferiority (Mani Lakshminarayanan)
Reporting/Tools (Mani Lakshminarayanan)
Joint Modeling (Larry Gould)
Program-wide Decision Making (Bin Yao/Karen Price)
Missing Data (Frank Liu/Stacy Lindborg)
Education (Fanni Natanegara/Mat Davis)
Joint with Pediatric community and ADSWG (Jasmina Savic)

8. DIA BSWG Impact
Published numerous papers, including Special Issue Pharmaceutical Statistics
Presented at several conferences, held webinars, taught short courses, etc.
Facilitated many productive conversations across academia/industry/regulatory
Joint DIA Bayes/AD conference (Feb 2015)

9. Bayesian Survey and Recommendations
© 2015 DIA, Inc. All rights reserved.

10. Bayesian Survey In 2012, we conducted an industry-wide Bayesian survey
first industry-wide survey to collect information on the use of Bayesian methods amongst statisticians working in medical product development
results and recommendations were submitted as a paper to be included in the Special Issue of Pharmaceutical Statistics

11. Implementation Hurdles
Insufficient knowledge of the Bayesian approach, particularly on the practical level
Lack of clarity of the regulatory position and/or lack of guidance and experience
Lack of tools including case examples and user-friendly software
Company-internal difficulties (lack of time, lack of support/guidance, general reluctance from team members to accept the Bayesian approach)
include difficulty in explaining Bayesian concepts and lack of understanding where it may be appropriate to consider Xx
Within an organization lack of precedents to motivate
Support from management and team

12. HOW TO IMPROVE THE PROGRESS ON BAYESIAN APPLICATION IN MEDICAL PRODUCT DEVELOPMENT?

13. Recommendation #1: Bayesian education
Familiarity
Understanding
Appreciation
Internal and external forums
Internal live training is preferred
Central repository of literature (books/articles) organized by topics to be made available
Learning process is continuous!
Through education, all stakeholders can be made familiar and have basic understanding of advantages and limitations of Bayesian methods

14. Recommendation #2: Centralized location for case examples
Searchable by categories e.g. statistical model, drug phase, therapeutic area
Include examples from internal decision making to regulatory submission
Be made available publicly or internally

15. Recommendation #3: Internal Bayesian infrastructure
A group of statisticians with dedicated time for focusing on Bayesian methods
allow systematic training of statisticians and non-statisticians
recruitment of statisticians with thorough Bayesian training
collaboration with other functions

16. Recommendation #4: Interaction between Bayesian and frequentist statisticians and other stakeholders
How to convince stakeholders that both Bayesian and frequentist methods are useful
Focus should be on the problem and not the methods
convey balanced opinions when presenting a Bayesian approach to a problem
Our focus in 2013/2014 via paper

17. Specific Actions
© 2015 DIA, Inc. All rights reserved.

18. General Needs
Acceptance of type 1 error control through simulation in AD and non-AD
Agreement to use informative priors
Greater and consistent knowledge of AD and Bayesian methods in Industry and Regulatory
Sponsors and agency can discuss and reach agreement on technical questions related to modeling and simulations earlier
Consistent format to enable easier agency review when sponsors submit simulations
More transparency from Sponsors on use of Bayes/AD
More insight and metrics in terms of the types of AD/Bayesian approaches which are submitted
Better use of modeling clinical endpoints
AD/Bayesian focus in one of the FDA-funded Centers of Excellence in Regulatory Science and Innovation.
This could allow the FDA to leverage external expertise in these areas in an academic setting which could better inform agency decisions.
Bayesian topics: type 1 error (define null space), agreement on informative priors as appropriate, consistent format,
Education is important!

19. Areas BSWG will focus moving forward
Case examples
Best practices document(s)
Enhanced medical involvement and outreach
Continue with existing subteams
© 2015 DIA, Inc. All rights reserved.