Transcript presentation - BMS and ANed

Biostatisticians lead the way
Kit Roes
Biostatistics UMC Utrecht
BMS-ANed 22oktober 2009
Biostatistics as conscience of the life sciences
François Chifflart (1825–1901), La Conscience (after Victor Hugo)
Conscience
 Is an ability or a faculty that distinguishes whether one's actions
are right or wrong.
 It leads to feelings of remorse when a human does things that
go against his/her moral values, and to feelings of rectitude or
integrity when actions conform to moral values.
 The extent to which such moral judgments are, or should be,
based wholly in reason has been a matter of controversy almost
throughout the history of Western philosophy.
Perspective(s)
(and potential conflicts of interests?)
This presentation
 Biostatistics in pharma industry
 Examples
 Data irregularities
 FDA Advisory Committee
 DSMB & stopping studies
 Roles & responsibilities: are statisticians sufficiently
equipped?
 Concluding remarks
Biostatistics in pharma R&D
Statistics / Quantitative modeling
Target
Lead
Lead
Pre-clinical / CMC
Optimization Development
I
Clinical
Development II a
Discovery
Research
Exploratory
Development
Pre-clinical / CMC
Marketing &
Registration Sales
Development
II b
Clinical
Development III
Full Development
and Launch
Development
Statistics Leadership
At R&D Leadership level
Early Stage
Pharmacology and toxicology
Biomarkers, Pharmacogenetics &omics
Bio-analysis & assay validation
Early clinical development
Collaborating scientists
Relatively unstructured
Scientists also analyze data
Statisticians more advanced
Quality control
Late Stage
Clinical Development
Integrated analyses for submissions
Interaction with regulatory authorities,
Phase IV, safety monitoring &
signal detection
and publication support
Clinical teams with clear
roles & responsibilities
Statisticians analyze data
Routinely involved in
interaction with authorities
Teams within statistics
collaborate on same project
Quality control
Pharmaceutical statisticians
Little Use of Statistics 
“Required” use Statistics 
Tactical use of Statistics 
Strategic use of Statistics
& “Statistical thinking” 
1955
2009
Rockhold, 2002
Regulated
 Confirmatory trials


ICH E9
Additional guidance (FDA, EMEA)


DMCs, Non-inferiority, Flexible designs, Missing data
……
 All registrations in US: re-analysis of key results, publicly available,
including critical assessments (Advisory Committees)
 EMEA: Critical assessment (including statistical issues), publicly
available.
 Obligation to report results of all analyses performed (also post-hoc)
Perfect world?
Data irregularities
 Electronic patient diaries (ePRO)
 Huge amounts of data
 Graphical checking as
initial data analysis
 Can reveal patterns
 With elaborate algorithm identified sites
suspect of clustering of entry times
 See also O’Gorman (DIA Clinical Forum,
2009)
As it should be…
Pattern emerges……
Data irregularities
Through internal Advisory Committee on
suspected fraud (with statistician):
 Sites investigated / audited / confirmed
 Reporting to authorities, including solution for
analysis
 Pro-active measures for all new trials with
similar devices
FDA Advisory Committee
 Independent experts
 Advise FDA on approval
 Questions



Efficacious in…
Safe in…
(Benefit / Risk?)
FDA Briefing
Company Briefing
Question and answer
 Vote
FDA Advisory Committee
http://www.fda.gov/downloads/AdvisoryCommittees/CommitteesMeetingMat
erials/Drugs/PsychopharmacologicDrugsAdvisoryCommittee/UCM179980.p
df
DR. SZEGEDI: With regard to suicidality, be it completed
suicide or suicide attempts, we have done several analyses and
I would like Dr. Kit Rous from Biometrix to explain them to you.
DR. RUES: Thank you, Dr. Szegedi.
If I understand your question correctly, I think the concern is that
actually the rates of suicide to suicide attempts might differ in
short and long term where actually as a result, the comparison
across all time versus placebo would not provide the correct
picture. We have addressed that question and we have done so
by actually comparing suicides and suicide attempts because
there are only very few suicides to draw any conclusions
comparatively for asenapine and olanzapine,
Topics of statistical nature
 Missing data / drop-out
 Survival analysis / hazard rate assumptions
 Treatment by Center and US vs Non-US
 Strength of evidence (positive/negative/failed
trials)
Key roles and challenges
 Prepare for and address the “statistical”
questions.
 Communication
 (Help) Ensure the whole team does not go
beyond what the data allow to conclude.
 Maintain integrity (while you cannot have
everything your way)
DSMB LIFT
The LIFT study (NEJM, Aug 14, 2008.The Effect of Tibolone in Older
Postmenopausal Women, S.R. Cummings et al.)
Randomized, placebo controlled study in 4538 osteoporotic,
postmenopausal women to evaluate the effect of 3 yrs treatment
for prevention of vertebral fractures.
 Primary endpoint: Incident vertebral factures after 3 years of
treatment
 Secondary: Clinical fractures
 Key safety (pre-defined, partly):
 Endometrial hyperplasia Breast cancer, All cancer
 CHD, Stroke or TIA
 VTE
DSMB Charter
 Membership, responsibilities, procedures, reports,
unblinding, conflict of interest.
 Key efficacy data and key safety data to be
monitored.
 Sequential monitoring of incidence of vertebral
fractures (Lan-Demets with a symmetric 5% O’BrienFleming-type spending function).
Additional considerations when interpreting the data:
“No single statistical decision rule or procedure can take the place of the
well reasoned consideration of all aspects of the data by a group of
concerned, competent and experienced persons with a wide range of
scientific backgrounds and points of view.”
(1): A safety concern
August 2005 (average follow up 2.4 years) the
LIFT DSMB:
 Indicates shortly before their August 26, 2005
meeting that they might share a safety
concern with representatives of Steering
Committee and Sponsor.
 That may involve (partial) unblinding.
Steering Cee & Sponsor procedure
Written and dated agreement was made to the following
procedure (statistician responsible):
Potential unblinding to full report if needed:
 Two key Steering Committee members (incl. chair).
 The (non voting) Steering Committee member on behalf
of the sponsor.
 Two senior managers from the sponsor (Medical Affairs
(MD) and statistician), not involved in trial.
Subsequently, a larger group could be unblinded to limited
information, if necessary for execution of the DSMB
recommendation.
(1): What happened?
DSMB recommends:
 Inform LIFT participants and scientific community of
observed increased risk of stroke associated with
tibolone in this study.
 Include rates, relative hazards, CI’s and p-values.
 Participants to re-consent.
 Continue the trial for important additional scientific data.
Steering Committee and sponsor representatives:
 Decided not to be further unblinded (beyond what will be
made public).
BMJ 2005, 8 October. LIFT study to continue as planned. D.E. Grobbee.
(2): Stopping the trial
February 2006 (6 months later)
 DSMB recommends discontinuation:
 Increased risk of stroke persisted in the LIFT study.
 Primary endpoint established, crossing the pre-defined
stopping boundary.
 Sponsor representatives and key Steering Committee
members enabled to review unblinded report (after
written confidentiality statement and over the weekend).
 Sponsor and Steering Committee reach same conclusion
and support DSMB recommendation.
BMJ 2006,18 March, LIFT study is discontinued. S.R. Cummings on behalf of LIFT
Steering Committee
DSMBs
Key area where statisticians continue to make a huge
contribution to clinical research integrity.
 Innovation in statistical methodology
 Impact on DSMB operation and status
 Typically area where statisticians are in the lead
(although not the chair)



Ellenberg, Fleming & De Mets, Whitehead, Pocock
EMEA Guideline, FDA Guidance
Internal pharma SOPs
Roles & responsibilities
Adequately equipped?
 Technical knowledge and continuous innovation
 Non-quantitative (biology, medicine, regulations)
 Collaboration and teamwork
 Communication
 Leadership
 Integrity (ethics….)
Unscientific: Google hits
 Chemistry and Ethics
 Physics and Ethics
 Mathematics and Ethics
 Statistics and Ethics
 Psychology and Ethics
 Medicine and Ethics
 Law and Ethics
5.5 M
6M
7M
12 M
15 M
19 M
40 M
Professional standards
 ISI Declaration of Professional Ethics
 ASA Ethical Guidelines for Statistical Practice
 RSS Code of Conduct
 Statistics Netherlands (CBS), and Norway,
and……
 ………..
ISI Declaration of Professional Ethics




Conflicting interests
Guarding against misuse and misinterpretation
Benefits as large as possible
Objective, transparent about limitations




Clarity in roles and responsibilities
Impartial assessment of alternative methods
No pre-emption of outcomes
Safeguard privileged info while revealing methods
Colleagues



Maintaining confidence in statistics
Transparency of methods
Knowing ones own ethical principles and those of
collaborators
Subjects

Principles of Informed Consent
Society
Paymasters
ASA and RSS
Add:
 Statistical analysis should be open to assessment,
limits and source of data analyzed visible.
 Data available for analysis by appropriate others
 Acknowledge that statistician can be overruled by
others (RSS)
Concluding remarks
 Keep our own conscience clear
 Peer review and collaborative work on projects*
 Innovate
 Integrity maintained as part of group shared value
 Organization and seniority to facilitate
 Educate ourselves (if not already done)
 Statistics and ethics (curricula available)
* Registration of VVS-Biostatisticians