Data Management Issues

1
Session 6

• Data Management Issues

2

• Its 3 months into the study and you have
  recruited only 2 patients. You were supposed to
  have recruited 12 patients by now. Your PO is
  threatening to stop the study. What can you do to
  save your study?

3
Collaborator Interest Curve
4
(No Transcript)
5
Encouragement

• Frequent (positive) updates
• Recruitment statistics
• Related publications
• Facilitation of activities
• Intermediate rewards
• Publications from the work
• Reimbursement for activities
• Peer pressure
• Recruitment statistics
• Commitment to recruit certain patients

6
Other Techniques That You Have Tried?
7
It is early in the course of the trial and you
want to plan your data quality control
procedures. You have initiated onsite monitoring.
Your DSMP calls for periodic queries of the data
to look for inconsistencies.
8
Available Data Elements

• Mayo score
• Bleeding score, stool frequency, physician global
  assessment, mucosal appearance (only wk 0 and 12)
• IBDQ quality of life measure
• Single question about disease activity
• Single question about well being
• Concomitant medications
• Adverse events
• Baseline demographics and medications

9
What strategy would you employ?
10
What We Did

• Automated range checks
• Double data entry
• Query drug data for prohibited medications
• Patient level correlations of disease activity
  measures looking for inconsistent results
• Disease activity improving and quality of life
  worsening (and vice versa)
• When identified would query clinical coordinator
  to confirm the data on CRFs

11
It is time for the first DSMB meeting. Data are
needed for the meeting. What should you provide
them? What role should the investigators have in
preparing these reports?
12
What To Provide (As Decided In the DSMB Charter)

• Monitoring reports
• Literature update
• Grouped data
• Enrollment data
• Overall by site
• Observed vs. expected to meet goal
• Demographics
• SAEs

13
What To Provide (As Decided In the DSMB Charter)

• Data stratified by arm
• Primary outcome
• Key secondary outcomes
• AEs
• SAEs

14
Should the DSMB Be Provided with Unblinded Data?

• Active vs. Placebo for effectiveness and AEs
• A vs. B for effectiveness and AEs
• A vs. B for effectiveness AND Active vs. Placebo
  for AEs
• A vs. B for effectiveness AND X vs. Y for AEs

15
Should the DSMB Be Provided with Unblinded Data?

• Could stop a trial early for effectiveness or
  safety reasons
• Difficult to interpret safety data without
  knowledge of treatment assignment
• Could stop for evidence of lack of effectiveness
  (i.e. futility), but require knowledge of
  treatment arm for that

16
Should the DSMB Be Provided with Unblinded Data?

• DSMBs should always have access to actual
  treatment assignments
• DSMBs are there to protect subjects
• Members have been chosen to be (relatively) free
  of conflicts of interest
• Why do we want to make it harder for them to do
  their job?
• Using different codes for efficacy and safety
  hinders risk/benefit assessment

17
Should the DSMB Be Provided with Unblinded Data?

• We went with
• A vs. B for effectiveness AND X vs. Y for AEs
• Separate sealed envelopes available to the DSMB
  Chair
• Kept the study statistician at least partially
  blinded to results

18
What Role Should PI/Investigators Play in
Preparing the Data
19
What Role Should PI/Investigators Play in
Preparing the Data

• Strategy A Hands off approach other than
  writing the DSMP
• Strategy B Hands on approach without unblinding
• Using pooled data, create all of the tables and
  figures that will be created for the final report
• Will pick up data inconsistencies, protocol
  violations, etc. that are likely to be missed
  with strategy A and are likely not included in
  other QA checks
• Will speed up final analyses at end of the study
• Risks biasing the investigator (correctly or
  incorrectly) if know the overall response rate

20
What Role Should PI/Investigators Play in
Preparing the Data

• Best practice prepare DSMB report templates
  before data are available
• If investigators will be able to review pooled
  data, no reason they cant be involved in
  reviewing tables beforehand
• Often useful to have an open DSMB session at
  which investigators can discuss study status with
  DSMB, including any concerns raised by pooled
  data (eg, event rate too low, may need to
  increase sample size)

21
What Role Should PI/Investigators Play in
Preparing the Data

• We used strategy A
• I would use strategy B next time

22
What About the Study Statistician?

• Statistician is preparing interim reports
• Statistician knows interim results
• Suppose
• Newly reported data from another study suggests
  primary endpoint is suboptimal
• Blinded study team proposes changing primary
  endpoint
• Statistician knows that data are looking very
  good on primary endpoint in this study, less good
  on proposed new endpoint

23
A Few Comments on Investigating Data
Inconsistencies

• We stored all CRFs off site from PI and PM (at
  request of the CRCU)
• PM had ability to query database
• Had PM been able to proof / QA check the CRFs
  prior to data entry it would have saved time
  later
• Resolve issues from comments on CRFs
• Need person who knows protocol well to catch
  inconsistencies, incorrect completion of forms

24
All recruitment is complete and all patients have
finished the protocol, what now?
25
Final Steps

• Final monitoring
• Final data queries
• Database lock
• Data analysis
• Final reports / manuscripts

26
Final Monitoring

• What to monitor?
• What about sites who have not enrolled since
  their last monitoring visit?

27
Final Monitoring

• Regulatory documents
• Drug supply
• Data elements
• Inclusion criteria
• Primary and key secondary outcomes
• Other

28
Final Monitoring

• What we did for sites with no activity since last
  visit
• Source documents had not been reviewed for key
  outcome variables for all patients and mistakes
  had been documented in earlier monitoring
• Copies of de-identified (other than study ID )
  source documents sent to DCC for review

29
Final Documentation

• How to be certain that you can document the
  source of the results for
• Drafting manuscript
• Revisions of manuscripts
• Other queries

30
Final Documentation

• A reasonable approach
• Create a cumulative results log
• Create a cumulative statistical code document
  with extensive documentation of the purpose of
  the code
• Decision log documenting all decisions
• Create a cumulative draft manuscript with more
  data than possibly will be included in the final
  reports and papers
• Every data point in the cumulative draft contains
  a comment field with source of result in the
  statistical code and/or cumulative results log

31
Final Thoughts from Participants