The Role of Statistics Post-Market

1
The Role of Statistics Post-Market

• Jay Herson, Ph.D.
• Johns Hopkins University
• Sept 29, 2006

2
Post-Market Trials

• Phase IV
• Label extension or modification
• Surveillance
• Investigator INDs
• Seeding Studies
• Biostatistician input much needed

3
Roles

• Role should not be described as a list of tasks
• Role of the biostatistician in post-market
  industry trials is to be an advocate for GOOD
  SCIENCE.

4
Wisdom

• Good Science is Good Business but
• Good Business is not necessarily good science
• Credible and timely results is good business in
  the long run

5
Good ScienceDesign Stage

• Objectives
• Primary and secondary endpoints
• Eligibility
• Power and precision
• Doses, regimens
• Control Group
• Non-inferiority trials

6
Good Science--Results

• Results available to patients and physicians
• Publication regardless of results
• Clear definition of original objectives of the
  trial
• Appropriate measures of precision, exposure
• Post-hoc power

7
Advocate

• Independent Review throughout the trial

8
Post Market Trials

• Then and Now

9
Large Trials

• GUSTOt-PA vs streptokinase
• ATLASadjuvant tamoxifen
• COMMITacute MI asparin vs clopidogrel
• CPPTlipids, cholestiramine
• CONCORDEHIV, AZT
• WHI

10
Surprising Results

• BHATpropranalol
• CASTencainide, flecainide

11
Cooperative Meta Analyses

• ATTantithrombolytics
• EBCTCGearly breast cancer

12
OncologyPractice Changing Trials

• Aromatase inhibatorsbreast cancer
• bevacizumab-VEGF, colorectal
• trastuzimabbreast cancer
• taxaneslung, ovarian
• platinumlung ovarian

13
However.

• Most of the trials just referenced were not
  conducted by industry

14
Industry Sponsorship

• Pre-Market---Science / Regulatory driven
• Post-MarketMarketing driven

15
Marketing Influence

• No trials undertaken whose results could threaten
  market share
• Results suppressed if they threaten market share

16
Industry Sponsorship

• Phase IVUnder control of marketing
• Anti-depressants19/21 trials head-to-head
  sponsors product superior.
• Low dose for competitor
• Different schedules
• Simple conditions
• Not double blind

17
The Non-Inferiority Complex

• Sloppy trials
• Result is under control of sponsor
• Delta
• Active control
• Sub-potent dose
• Terminate before effects
• Imprecise measurement

18
Industry Sponsorship

• Publication bias
• Results not matched with original objectives
• Cox-2 inhibitor trials / bureaucracy

19
Industry Sponsorship

• Advertising agency purchased CRO
• Seeding studies
• Design problems
• Organizational problems

20
Good News

• William Olson, Ortho-McNeil
• Patent 6,339,105
• A regimen for the administration of tramadol for
  the treatment of analgesia
• Slower titration rate
• Reduction in side effects

21
Opportunities

• Bayesian design and analysis
• Testing groundsadaptive designs
• Meta analysis
• Methodology
• Fixed vs Random Effects
• Mixed models

22
A. MugglinMedical Devices

• Fine illustration of the role of statisticians
• Interesting SCD-HeFT cost-effectiveness study
• How do results of this information get
  disseminated to stakeholderspatients and their
  caregivers?
• Should not be used as sole justification for
  higher prices for the product.

23
Mugglin--Future

• Globalization
• Aging of populations
• Cost-effectiveness
• IOM recommendation
• Genetics

24
Genetics

• Revisit approved but unpopular drugs due to
  knowledge of what DNA is associated with
  response, low toxicity.
• bucindolol vs other beta blockers for heart
  failure
• tamoxifen vs aromatase inhibitors in breast
  cancer

25
Gould--Surveillance

• fialuridine (FIAU), 1993
• Good list of limitations (Considerations and
  Issues)
• Accumulating information of RR and the lower 5
  point---how long to watch a signal?
• Retrospective pharmacovigilence

26
Gould--Bayes

• Definitely a role for Bayesian methods here. The
  sooner the better.
• Empirical Bayes a good start
• Sponsor safety database
• OncologyMedDRA vs CTC 3.0

27
Gould--Contribution

• Shows the contribution a statistician can make to
  a difficult problem by thinking through the
  problem
• Hope for a common ground between plaintiffs and
  industry
• To deny the implications of this type of analysis
  is to deny our ability to reason

28
Marinac-Dabic FDA/CDRH

• Sources of data
• Areas requiring attention
• Motivate good science
• Post-market surveillance for the elderly
• Passive reporting system for high quality
  post-market surveillance data

29

• Thank You