Setting up a Data Safety Monitoring Board

1. Setting up a Data Safety Monitoring Board ASENT Meeting March 6, 2008 Jennifer Schumi, PhD Statistics Collaborative, Inc. jennifer@statcollab.com

2. Five steps to setting up a DSMB • Decide why you need one • Choose the Board and its chairman • Choose the reporting statistician • Define how the Board operates • Write its charter

3. What is a D(ata)S(afety)M(onitoring)B(oard)? • A committee charged with monitoring • safety • efficacy • progress of a clinical trial • Aka DMC, IDMC, …

4. Rationale for using DSMBs in research • Ethical compact protecting trial participants • Sponsor: regulatory responsibilities for reporting financial incentive to end trial early • May also advise about changes in protocol, procedures • NIH often uses DSMB in an advisory capacity – different from industry-sponsored trials

5. 1. Define why you need an external DSMB • Scope of advice and decisions • Broad: Safety and efficacy and data quality • More limited: safety only • Who makes final decisions for each area

6. Standards for clinical trials • NIH Clinical Trial Committee Guide, 1979 “every clinical trial should have provision for data and safety monitoring… should be commensurate with risks” • During the 1980’s Many models developed; vary across U.S. & Europe • 1994 NIH Committee on Clinical Trial Monitoring: “all trials, even those that pose little likelihood of harm, should consider an external monitoring body.”

7. Standards for clinical trials, continued • 2006 FDA guidance document: Establishment and Operation of Clinical Trial Data Monitoring Committees “Full employment act” for statisticians Describes in detail… need for independence of DMC possible models for reporting statistician role of study team, interaction with sponsor

8. Trials that need a DSMB • Double-blind • Large (hundreds, thousands of subjects) • Multi-center/multi-national • Long duration • Endpoint: death or stroke or …

9. Trials that need a DSMB, cont’d • Participants have high intrinsic mortality risk • HIV infection, cancer • Sepsis, pulmonary disease, cardiac failure • Trial studying a new chemical entity • Recommended (strongly) by regulatory agency

10. Trials that DON’T need a DSMB • Phase I studies, pilot studies (some) • Studies of symptom relief • Studies with other very close safety monitoring • Timeline so short the DSMB can’t operate

11. NIH-type trial

12. Industry trial

13. Monitoring safety data • Serious adverse events (in cancer, Grades 3 and 4) • Adverse events (severity, related to drug) • Laboratory data liver function tests, hematology, changes over time • Disease/treatment specific scales • Discontinuation rates

14. Who watches safety? And how? • Investigators – individual patients • Pharmacovigilance – SAEs, case by case, across study • IRBs – SAEs at individual study sites • Study team (sponsor) – SAEs, other data, individual cases pooled across all treatment groups • DSMB – SAEs, other data, aggregated by treatment group

15. Problems with knowing unblinded interim results • New science, finances may cause want to change: • primary endpoint • entry criteria, evaluable population • concomitant medications • size of the trial • Not appropriate if proposer of change knows interim results

16. Roles of DSMB – safety and efficacy • Establish statistical boundaries for efficacy and futility • Statistical issues for efficacy settled • Statistical issues for futility in flux but reasonably settled • Type I error not the most important thing to protect

17. Roles of DSMB-expanded • Safety • Efficacy • Study and scientific integrity • Review protocol, CRFs • Review procedures that would affect data quality • Review quality & integrity of study data • Monitor accrual & drop out, compliance

18. 2. Members and chairman • Voting • Physician(s) in specialty area (disease, side effects) • Epidemiologist/trial methodologist • Statistician • Clinical pharmacologist/safety specialist? • Ethicist, patient representative, lawyer? • Need effective chairman

19. 2. Members and chairman • Non-Voting • Study or steering committee chair • Sponsor representatives • Reporting statistician

20. Voting members • 3-10 experts in disease, study drug, clinical trials • Multidisciplinary, independent • Disinterested – no conflict of interest • Experience on other DSMBs • Chair & statistician • Some inexperienced to train them • Must take responsibilities seriously

21. 3. Choose the reporter • NIH – someone from the coordinating center • Industry – • In-house statistician • Contract research organization • Independent statistical group • Statistician on the DSMB • Independence  ignorance

22. 4. Define the operations • Data • “Cleanliness is next to godliness” Not for DSMB!!! • Timeliness more important than cleanliness • Report must aim for clarity and focus • Structure of meetings

23. Meetings • (Brief executive session) • Open session • Closed session • Executive session • Disseminate recommendations Open session Directly to sponsor representative

24. Purpose of open session • Update on progress of the trial • Report from the sponsor, Steering Committee • Relevant information from related trials • Open and honest discussions • Sponsor, investigators, and DSMB can share concerns

25. Participants in closed session • The DSMB only? + the statistician(s) preparing the report? + the Chair of the Steering Committee? + the sponsor? • Executive session – DSMB only

26. Contents and intent of report Purpose –allow DSMB to make informed decisions • Summary of protocol and outstanding issues • Recruitment and follow-up • Baseline data • Check of randomization • Timeliness of data & adjudication of endpoints • Adverse events with study-specific coding • Dosage of study medication • Vital signs and laboratory parameters • Outcome data

27. Reports to the DSMB: comments • Should not just be a subset of final tables • Purpose of interim monitoring  purpose of final analysis • Changing the study during the trial  what does the study show • Don’t be locked into rigid rules! • Data should NOT be blind (not everyone agrees)

28. Recommendations from the DSMB • Shared with Sponsor, Steering Committee, IRBs • Must prevent unblinding of study team • Be careful with communications! • During the trial, everyone reads tea leaves • DSMB must keep impeccable records • What did they know and when did they know it? • Did they change their behavior and rules in response to data?

29. Monitoring for safety • Searching for the unknown • Rare and often unexpected events • Extreme problem of multiplicity “I make no mockery of honest ad hockery”- I. J. Good

30. What type of safety bounds? • None – just rely on the judgment of the DSMB • Futility bounds for efficacy Symmetric – as hard to call unsafe as to call efficacy Asymmetric – less stringent than for efficacy we do not want to prove harm • Careful balance of risks and benefits

31. Taxonomy of adverse events • Expected events – balance risk to benefits • Unexpected, but not serious • Unexpected, serious • Unexpected, very serious • Not credible, but scary if true

32. Enhancing safety data • Spontaneous reports notoriously ambiguous • Formal data collection • Diaries • Endpoint committees • Ongoing reporting of prognostic factors • Special data collection forms

33. Reporting safety data • Don’t rely on coding systems • Classify and reclassify • Look at relevant lab data • Means (remember the CLT!) • High percentiles (but not min and max) • Outliers for very rare events

34. 5. Write the Charter • Mission and goals • Responsibilities (confidentiality, lack of conflict) • Communications by DSMB • Governance: chair, minutes, quorum, tie votes • Frequency of meetings • Requirement for decision rules, statistical guidelines • Obtaining data for review

35. Issues to hammer out • Blind, unblind, or partially blind • Sharing data with other DSMBs • Participating on related committees

36. Stopping a trial • Early stopping is a major decision • Can hurt patients/product if stopping delayed • Can doom product if stopping premature • DSMB must aim for defensible decisions • Once recommended – very hard to change

37. After the door is open • Full and honest disclosure of what happened

38. Recommendations to sponsor • Sponsor and DSMB must trust each other • Must choose DSMB on basis of expertise • Membership is not a “reward” for good recruitment in other trials • Structure of the meeting should not impede the ability of the DSMB to do its work • Always have face-to-face meetings when the DSMB is thinking about stopping

39. Recommendations to investigators • Need careful safety monitoring plan • Investigators in a multi-center trial should look at the monitoring plan

40. Suggestions from other DSMBs • Should be multidisciplinary • Should be committed to the trial • Meet in person when possible • Reporting group must understand trial and data • DSMB must have the right • to ask for additional presentations without telling sponsor • to hold executive sessions • to request/recommend additional expertise

41. Questions?

42. Informed consent • Agreement describing risks and benefits • People don’t enter trials to prove intervention is harmful