CSRC Think Tank: Real World Data to Assess Cardiovascular Safety- Can We Improve Efficiencies in Phase 3 Development?

Objectives: Evaluate the use of “real-world” data sources to provide sufficient and acceptable evidence to make informed decisions regarding CV safety risks of drugs.  What are the data? Methodologies? MACE definitions? Sources of bias? Analytical approaches? How can limitations be overcome?

7:30-8:00 Continental Breakfast

8:00am- 8:15am CSRC Welcome & Introduction- Mary Jane Geiger, MD (Regeneron)

8:15am-8:45 Session I: Background/Overview of Key Issues

8:45am-10:30am Session II: Data Types – Quality, Event Ascertainment, Limitations- Strengths, limitations, MACE (CV death, heart failure, non-fatal MI and stroke events) ascertainment /issues. Provide example(s) where data type has been used to identify a CV event.

Moderator: Philip Sager, MD (Stanford University)

  • Prospective Cohort Studies/Observational registries
  • Clinician reported data – Sebastian Schneeweiss, SD, MD (Harvard Medical School )(10 min)
  • Healthcare databases
  • Electronic health records (EHR) – Christina Mack, PhD (QuintilesIMS)(10 min)
  • Claims databases– Sharon-Lise Normand, PhD (Harvard Medical School)(10 min)
  • Linked databases: EHR/observational registries/claims databases – Nancy Dreyer, PhD (QuintilesIMS)(10 min)
  • How are data from different sources or within a data set with multiple entries linked? How might linkage introduce bias? What advantages do hybrid databases offer?

Discussion (55min)

What is the best source of evidence? Does an optimal data type exist? If not, what would this look like?

10:30am-10:45am: Break

10:45am-11:00am Session III: Casualty

  • How is causality determined? Challenges to interpretation? Impact of confounders? How much could bias explain observations? – Timothy Lash, DSc (Emory University)(15 min)

11:00am-12:50am Session IV: Study Design/Analytic Methods, MACE Definitions/Identification & Exposure.

Moderator: Jonathan Seltzer, MD (ACI Clinical)

  • Optimal methods approaches in Existing Data– Patrick Ryan, PhD (Johnson &Johnson)(10 min)
  • Pragmatic Trial Designs – capturing endpoints and integrating data from non-linked sources – Matthew Roe, MD (Duke University Medical Center)(10 min)
  • The Salford Lung Study: an example of an executed pragmatic trial – what can we learn? – Frank Rockhold, PhD (Duke Clinical Research Institute) (10 min)
  • CV Event Definitions – What endpoints can be assessed? What definitions are / should be used? Can they be applied uniformly across data types (eg, EMR vs claims)? What are potential confounders and how can these be managed? Does adjudication add merit?– Brad Hammill, DrPH (Duke Medical Center)(10 min)
  • Reliability of Data Sources to Ascertain Drug Use, Compliance, & Exposure – Alan Brookhart, PhD (University of NC)(10 min)

Discussion (60 min)

Jonathan Seltzer, MD (ACI Clinical)

What is an optimal study design/analytic method to minimize bias? Can/should MACE event definitions be the same in observational trials vs RCT?  What level of uncertainty is acceptable? Is consistency or uniformity of observational study results necessary or something other than trials?

12:50pm-1:30pm Lunch

1:30pm-3:15pm Session V: Methodologies to Reduce Bias and Estimate Effect Size.

Moderators: Mary Jane Geiger, MD, PhD (Regeneron)

  • Provide case examples for illustration. Bias: Sources of and methods to reduce bias. Does lack of randomization introduce bias? Brian Bradbury, DSc (Amgen)(15Min)
  • Analytical approaches for estimating effect size and interpreting results – Marc Suchard, MD, PhD (UCLA)(20 min)
    • Novel approaches
    • Reproducibility of results across multiple databases
  • Limitations of Big data & Relationship between Elements of Data –- Anita Pramoda (Owned Outcomes) (10 min)

Discussion (60min)

What level of evidence is needed to inform decision-making- regulatory, clinical practice? What is an acceptable odds ratio?

3:15pm-3:45pm Session VI: Key Points of Consensus & Recommendations.

Based on what is known about the current data, definitions, and methods, what are the minimum factors/considerations that would allow us to draw meaningful conclusions and believe them? What CV event definitions should be used?

Moderators:  Mary Jane Geiger, MD, PhD (Regeneron) Jonathan Seltzer, MD (ACI Clinical), Philip Sager, MD (Stanford University)


3:45pm- 4:00pm Wrap-up & Next Steps