Clinical Trial Design and other Statistical Issues

1
Clinical Trial Design and other Statistical Issues

• Mary A. Foulkes, Ph.D.
• Office of Biostatistics and Epidemiology

Blood Products Advisory Committee March 17, 2005
2
Chance favors the prepared mind Louis Pasteur
3
Novel Statistical Applications

• Usual approach
• Develop theory/model/method
• Seek application
• Critical Path approach
• No prior stat approach
• or existing approximation Develop new or
  improved theory/model/method

4
Quantitative Methods

• Maximizing efficiency while maintaining
  reliability
• Improved analytic approaches
• Flexible study designs
• Transparency
• Best Practices
• Underlying assumptions

5
CBER Products

• Blood, blood products and tissues
• Life-saving therapies
• Potential to transmit serious disease
• Cellular and gene therapies
• Vaccines

6
Study Endpoints

• Biomarkers
• Intermediate endpoints
• Composite endpoints
• Multiple endpoints
• Assessment scales

7
Statistics in Manufacturing

• Quality control in blood collection
• Low volume processing
• False positives highly burdensome
• Efficiency potential in alternative statistical
  approaches
• Vaccine lot consistency

8
Flexible Study Design

• Initial parameters may be inaccurate
• Interim modifications may improve trial ability
  to answer question
• Traditional approaches
• Group sequential designs with early stopping for
  efficacy, harm, futility
• Other changes if not influenced by knowledge of
  interim data
• Emerging approaches

9
Flexible Designs for Biologics

• CBER regulates cutting-edge products
• May be less information on design parameters,
  more need for flexibility
• Safety concerns
• Optimal design approaches?

10
Trial Design Analyses

• Non-inferiority trials
• ICH E10 gives genl principles, not specifics
• Optimal approach for defining acceptable
  difference not yet defined
• Missing data
• No preferred analytic approach
• Implications of assumptions

11
Summary

• Quantitative methods are central to many critical
  path issues
• Quantitative methods will be needed to evaluate
  potential gain from new approaches
• Statisticians and Epidemiologists involved in
  many issues identified by stakeholders