DMC Issues from a Pharmaceutical Industry Perspective
Steven Snapinn
Amgen
FDA-Industry Workshop
September 15, 2005
Outline
• Assessing the Need for a DMC
• Independence of the DMC
• Scope of the DMC’s Responsibilities
• Issues in Setting the Stopping Boundaries
• How Are Decisions Made?– Case Study 1: PRISM-PLUS
• Stopping for Futility– Case Study 2: CONSENSUS II
Assessing the Need for a DMC
• Considerations– Seriousness of the Medical Condition– Uncertainty of Efficacy and Safety– Size of the Trial– Duration of the Trial
• Benefits of Having a DMC– Credibility– Experience– Neutrality
Internal vs. External DMCs
• Early Development– Safety Monitoring by Study Team
• Dose Escalation Decisions
– Efficacy Monitoring by Internal DMC• Independent of Study Team
• Phase III– External DMC Typical
Independence of the DMC
• Committee Membership– DMC Independent of Sponsor– No Sponsor Participation in Closed
Session
• Preparation of Interim Reports– Regulatory Guidance Frowns on This
Being a Sponsor Responsibility– Sponsors Typically Contract
Independent Group
The Case for an Unblinded Sponsor Statistician
• “Firewall”• Familiarity with the Study• Control of the Allocation Schedule
and Interim Results– Sponsor’s Risk– Quality Assurance– Data Leaks
• What Constitutes “Independence”?
Scope of the DMC’s Responsibilities
• Safety Monitoring• Efficacy Monitoring• Timeliness and Accuracy of the
Database• Protocol Adherence?• Sample Size Re-estimation?• Requests for ad hoc Analyses
Issues in Setting the Stopping Boundaries
• Stopping Boundaries for Safety
• Stopping Boundaries for Efficacy: Protect Patients in the Trial or Protect All Patients?– Require Overwhelming Evidence or
Moderate Evidence?
– Are Patients Fully Informed?
Issues in Setting the Stopping Boundaries (continued)
• Review of Efficacy Data for Risk/Benefit Assessment Only– Set Extreme Efficacy Criterion (eg,
0.0005)
– Protection for Sponsor
• Group Sequential vs. Conditional Probability
Other Issues
• Partial vs. Full Unblinding
• Information Sharing Across DMCs
• Monitoring Noninferiority Trials
How Are Decisions Made?
• Decision-Making Authority– Independent Steering Committee– Sponsor– Sponsor’s Awkward Position
• Reclaiming Type I Error– Efficacy Boundary Crossed But Trial
Continues– Futility Boundaries
Case Study: PRISM-PLUS
• Patients with Acute Coronary Syndrome– Non-Q-Wave Myocardial Infarction
– Unstable Angina Pectoris
• Evaluation of Tirofiban, an Inhibitor of Platelet Aggregation
PRISM-PLUS Study Design
• Three Treatment Arms– Control Arm: Heparin Alone– Monotherapy Arm: Tirofiban Alone– Combination Arm: Heparin+Tirofiban
• Composite Endpoint– Refractory Ischemia/Readmission for
UAP– Myocardial Infarction– Death
Study Organization• Oversight by an Independent Steering
Committee– Sponsor Representatives Attend SC
Meetings– Makes Decisions on DMC
Recommendations
• Data Monitored by an Independent DMC– No Sponsor Representative (Other than
Unblinded Statistician)– Recommendations on Trial Modification to
SC
Interim Results of PRISM-PLUSComposite Endpoint
Time Point Heparin (N=351)
Tirofiban (N=345)
Combination (N=336)
2 Days 24 (6.8%) 26 (7.5%) 19 (5.6%)
7 Days 59 (16.8%) 59 (17.1%) 39 (11.6%)
30 Days 78 (22.2%) 81 (23.5%) 63 (18.8%)
Interim Results of PRISM-PLUSDeath
Time Point Heparin (N=351)
Tirofiban (N=345)
Combination (N=336)
2 Days 1 (0.3%) 2 (0.6%) 0 (0.0%)
7 Days 4 (1.1%) 16 (4.6%) 5 (1.5%)
30 Days 14 (4.0%) 21 (6.1%) 7 (2.1%)
Summary of Results
• Composite Endpoint Rates Similar in Heparin and Tirofiban Groups
• Death Rate Higher in the Tirofiban Group
• 7-Day Mortality– 16/345 vs. 4/351 - p-value = 0.006
• Pooling Heparin and Combo Groups– 16/345 vs. 9/687 - p-value = 0.001
• Is This Sufficient Evidence?
The Pharmaceutical Sponsor’s Dilemma
• DSMB Recommended Discontinuation of the Tirofiban Arm
• Steering Committee Felt That the Evidence Was Insufficient
• Representatives of the Sponsor Were Present at the Meeting
• Can the Sponsor Allow Randomization to Continue when the DMC Believes That Patients Will Die Unnecessarily?
Final Results of PRISMDeath
Time Point Heparin (N=1616)
Tirofiban (N=1616)
2 Days 4 (0.2%) 6 (0.4%)
7 Days 25 (1.6%) 16 (1.0%)
30 Days 59 (3.6%) 37 (2.3%)
Stopping for Futility
• Inability of a Trial to Meet Its Objectives– Operational vs. Statistical– Goal Is to Preserve Resources
• No Ethical Imperative• Group Sequential vs. Conditional
Probability– Frequentist vs. Bayesian Methods
Stopping for Futility (continued)
• One-Sided vs. Two-Sided Boundaries
• Can Alpha Be “Reclaimed”?
• Cost in Power and Secondary Objectives
• Need for Prior Agreement
Case Study: CONSENSUS II
• 9000 Patients with Acute Myocardial Infarction
• Comparison of Enalapril and Placebo
• Primary Endpoint is 6-Month Mortality
• Assumed Rates: 12.0% vs. 9.6%
CONSENSUS II Interim Monitoring
• Key Design Features– Frequent Interim Analyses– Terminate If Interim Results Clearly
Indicate Lack of Benefit
• Monitoring Plan Based on Conditional Probability– Future Rates Assumed Between Current
Rates and Originally-Assumed Rates– Alpha Reclaimed
CONSENSUS II Interim Results
Analysis Enal PBO Enal PBO Rej. Acc.
1st 40/6725.9%
31/6334.9%
None None
6th 178/20798.6%
159/20827.6%
38/35610.7%
34/3519.7%
1.1% 16.5%
7th 242/26909.0%
215/26938.0%
89/74412.0%
74/72410.2%
0.1% 58.4%
Final 312/304410.2%
286/30469.4%
164/147511.1%
146/14779.9%
0.01% 94.8%
All Rand. Pts Pts w/6 Mo. FU Cond. Prob
CONSENSUS II Final Results
0
0.02
0.04
0.06
0.08
0.1
0.12
1 31 61 91 121 151
Study Day
Pro
po
rtio
n D
ea
d.
Placebo
Enalapril
CONSENSUS II Lessons Learned• Stopping for futility is a difficult
problem– Without Clear Trend Toward Efficacy
or Harm There’s No Ethical Imperative
– Without Hope for Benefit Patients Should Not Be Subjected to Risks
– Prior Agreement Required• Within DMC• Between DMC and Trial Leadership
CONSENSUS II Lessons Learned (continued)
• Stopping Boundary– Flexibility Is an Advantage
– Reclaiming Alpha May Not Be Appropriate
– Basing Conditional Probabilities Only on Patients With Complete Follow-Up Was a Disadvantage