Using an Independent Statistician to Support a Data Monitoring Committee

1
Using an Independent Statistician to Support a
Data Monitoring Committee

• Patrick D. OMeara, Ph.D.
• Pat OMeara Associates, Inc.
• pat_at_patomeara.com

FDA/Industry Workshop 14-16 September 2005
2
Outline

• Introduction
• Checklist
• 2 examples
• Recommendations

3
Introduction
. . . the integrity of the trial is best
protected when the statistician preparing
unblinded data for the DMC is external to the
sponsor, especially for the critical studies
intended to provide definitive evidence of
effectiveness. - Draft Guidance On the
Establishment and Operation of Clinical Trial
Data Monitoring Committees
4
Introduction
Theres more to protecting the integrity of the
study than just engaging an independent
statistician to perform a survival analysis for
the DSMB.
5
Checklist

• Charter
• Contract
• Documentation from Sponsor
• Database
• Support from the Sponsor

6
DSMB Charter

• Independent statisticians role
• Who directs the IS work?
• Sufficient freedom to create tables/analyses
  requested by the DSMB
• In effect the IS acts as an employee of the DSMB
• Roles of other organizations who provide data
• CRO
• IVRS
• Central clinical lab
• Biomarker lab

7
Contract

• Tasks assigned to the IS
• Programming of the tables, graphs, listings
• Who will write them?
• How will they be verified ON THE ISs SYSTEM?
• If written by the Sponsor, what changes are
  allowed?
• Who is in responsible for the data?
• Especially important if a CRO has been retained
  to monitor the study and to prepare the data for
  interim analysis
• Adding variables to an analysis database
• Correcting errors, inconsistencies
• Interpretation

8
Contract

• Time to assemble the report
• From time IS receives the data
• Once agreed Sponsor must not let it slide.
• Time to prepare a presentation
• Especially important at formal interim analysis
  when auxiliary analyses may be needed to support
  result.

9
Documentation from Sponsor

• Protocol blank CRF
• Informed consent forms
• Formal Statistical Analysis Plan
• Includes a clear statement of decision rules for
  any interim analysis
• Null and Alternative Hypotheses
• Significance level
• Test statistics and methods
• Futility, efficacy
• Adjudication process rules
• Investigators Brochure

10
Documentation from Sponsor

• Proposed list of Tables, Graphs, Listings
• Any special consideration in the studies
• The definition and processing of SAEs in a
  clinical endpoint study
• For example, stroke
• Biomarker lab data

11
Database

• Sources of the data
• CRF database (SAS)
• Analysis database (SAS)
• Clinical laboratory (SAS, text)
• IVRS (SAS, text, or Excel)
• Biomarker or specialty laboratory (text, Excel)
• SAE database (expedited review pharmacovig.)
• Current death list (Excel)
• 24-hour reports of SAE/Clinical endpoints

12
Database

• Detailed specifications derivations of derived
  variable
• Annotated CRF
• Description of special processes
• Topics that Sponsor and IS should discuss
• Frequency of updates
• Timing before reports
• Robustness of interim cuts of the database
• Do AEs disappear?
• Are some data sources more reliable or current
  than others?

13
Support from the Sponsor

• Availability by telephone or email of
• A Statistician who can speak with authority about
  the study and proposed analysis.
• A Statistician/Data analyst/Programmer who can
  answer detailed questions about the data.
• Face-to-face meeting with key project team
  members data management and statistics
• Learn the system that produces the data for DSMB
• Especially useful when resolving inconsistencies.

14
Example 1

• Two treatments Planned subjects 2200
• 28-day all-cause mortality
• Two interim analyses planned
• Protocol, SAP
• Tables, listings from Sponsors standard library
• The Job IS to reproduce TLG and present
• Randomization schedule from IVRS group
• 4 data sets for analysis with detailed
  specification
• 8 days to prepare and ship report

15
Example 1

• Statistician assigned to database quality
• Sponsors project statistician provided SAS code
  to implement formal interim analysis.
• A dummy r.s. using A odd, B even
• 4-5 test shipments of data before 1st interim
• Using last test shipment, 100 check of all
  tables against set produced by Sponsor.

16
Example 1

• 1st Interim Report and Analysis
• Timeline for report was squeezed by 1-2 days
• DSMB decision to continue study without change
• 2nd Interim Report and Analysis
• Efficacy was demonstrated.
• Auxiliary analyses demonstrated consistent trends
  across many subgroups
• DSMB recommended stopping for efficacy
• IS presented results to Sponsor executive comm.
• Study stopped.

17
Example 1

• Lessons learned
• Well-defined roles and responsibilities
  contributed to team environment.
• Extra data transfers allowed practice so that
  final transfers went smoothly and tight timelines
  could be met.
• Time spent reviewing database was a big
  contributor to success of the project.
• 100 check of tables using dummy r.s. was
  essential.

18
Example 2

• Two treatments 1800 patients
• All-cause mortality
• DSMB meetings every 3 months at beginning
• Protocol, no SAP
• Review imbalance every 8 deaths for 1st 100
• IVRS, Sponsors DM, Biomarker, Clinical lab
• Monthly updates of clinical database via FTP
• Sponsor provided SAS programs for data from
  Clinical database (primary motivation )

19
Example 2

• Programs designed for VAX local system PC.
• File references in every program had to be
  changed.
• Each program contained an extensive block of code
  to merge in randomization schedule.
• Logic errors in several programs.
• After a month, client resent programs with
  changes but the changes were not documented and
  all the file references had to be changed again.

20
Example 2

• Meanwhile, Biomarker data had been coded by the
  lab to prevent inadvertent blinding.
• Coded results were manually transcribed into
  database no source record of original value.
• Coded with C algorithm using radix-32.
• When decoded found that there had been many
  transcription errors --- invalid values.

21
Example 2

• DSMB expressed concern about decision rule for
  stopping for safety.
• Requested Monte Carlo study.
• Decided to meet again in 10 days to discuss.
• Monte Carlo showed that under reasonable
  assumptions, chances of stopping were less than
  .05.
• Recommended changing rule so Pr(Stopping) .15.
• Next DSMB, no improvement in safety study
  stopped.

22
Example 2

• Lessons learned
• Transferring SAS programs between systems
  requires careful planning and lots of work.
• Need to specify who is responsible for correct
  implementation on local system.
• Nothing in contract that said IS could do
  additional analyses suggested by DSMB.
• Serious decisions by DSMB could result in
  liability.
• Unforeseen problems can heavily influence the
  amount of time spent on a project.

23
Recommendations

• Provide the independent statistician with all the
  available information about the study
• Protocol
• Statistical Analysis Plan
• Assign project team members to with the IS.
• Charter should state IS works for the DSMB.
• IF Sponsor decides to provide programs,
• Work with IS when designing
• State in contract charter who is responsible
• Contract should indemnify IS.

24
The End