The Role of the Data Monitoring Committee

Society for Clinical Research Associates

Philadelphia, PA

April 23, 2010

Role of Data Monitoring Committees (DMCs)

Why use a DMC? Roles, Objectives and Setup for DMCs Operational considerations (Brief) Statistical considerations: p-value

adjustments Where do companies get it wrong:

Administrative analyses: look without adjustment Use DMCs for safety only Blinding the DMC

Reading List

Guidance for Clinical Trial Sponsors

http://www.fda.gov/cber/guidelines.htm

Fundamentals of Clinical Trials

Friedman, Furberg, DeMets

Polio: An American Story

David Oshinsky

Reasons for DMCs

Ethics Insurance

Reasons for DMCs: Ethics

Oversight function to insure patient safety Monitor and be able to quickly react to any

untoward safety events

Reasons for DMC: Insurance

Operational Issue Provide oversight function of study progress Insure study will have reasonable likelihood of

achieving basic objectives. Early warning system for operational issues

Critical design parameters• Expected treatment difference, control response

rate, variance of primary endpoint Stop early

• No efficacy or unacceptable safety• Early, compelling, untoward, unexpected efficacy

Committees

Data Monitoring Committee Executive Committee Others

Data Analysis Center Adjudication Committee Central Labs

DMC

Review data tabulations from ongoing clinical trial

Deliberations are confidential Make recommendations to Exec.

Committee

DMC Membership

Independent clinical members Therapeutic area experts No vested interest in company or outcome of

trial Independent statistician Ethicist (optional)

No Involvement in DMCs

Company/Sponsor/Client Investigators participating in the study The investigator’s contact at company Any individual who can change or influence the

recruitment of patients Individuals with data classification

responsibilities Individuals who could control or change study

design, objectives or planned analyses

Executive Committee/Steering Committee

Responsible for study oversight and conduct

Membership: company, investigators Decision makers

Key Point

DMC Independent Access to data No decisions, recommendations to Exec Comm.

Exec Committee/Steering Committee Company representation Investigator representation (optional) NO access to data Decision makers, based on recommendations from

DMC

DMC Charter

Define roles and responsibilities of DMC members

Communication with Executive Committee Structure and timing of meeting Scope of data reviews

DMC Meetings

Open sessions Executive Committee DMC No unblinded data

Closed sessions Unblinded data review DMC Only

Minutes Document deliberations Confidentiality is key

Example: Centocor

Centoxin: potential blockbuster with estimated $1B/year market potential

NYT: 12 Feb 1993 Centoxin

Efficacy endpoint in pivotal study changed based on knowledge of interim results

NDA terminated

Operational Considerations

Keep review scope focused (“interim analysis” is not final analysis)

Emphasize simple tables and graphics, not listings

Information needs to be Current Current is more important than clean

Plan for Rapid Retrieval of Outcomes

Important to minimize time lag between CRF at site vs inhouse

Paper Short forms, worksheets Phone calls Working/temporary databases Help Desk support

EDC very helpful alternative to paper

Statistical Issue

When DMC looks at interim data, p-value adjustments are necessary Avoid over reaction to early trends Maintain nominal alpha level of 0.05 for the

final analysis

Goal: Control type I error and maintain nominal 0.05 alpha for

final analysis Need to set a high statistical bar for

Interim Looks Interim analyses utilize p-values at levels

of approx. 0.0001 at each look vs 0.05 at each look

E.g., 6 interim analyses (6 looks):

final alpha = 0.05 – 6*0.0001 = 0.0494

Key Point

There are sound reasons to have a DMC monitor ongoing data

P-value adjustments need to be made Estimates of efficacy and safety effects are

based on small sample sizes Possibility exists to overreact to early trends

P-value adjustments for interim looks can be very small and final alpha can be maintained very near 0.05

Interim survival analyses comparing mortality in clofibrate and placebo-treated participants in the Coronary Drug Project. A positive Z value favors placebo.

Where Companies Get it Wrong

Administrative analyses DMC for safety only Blinding the DMC

“Administrative Analyses”

Look at data with no intent to modify study Look at data for operational (“insurance”)

issues Since no intent to change, no adjustment

of p-values should be necessary If efficacy data are involved, adjustment is

needed Always a potential to overreact to early trends Major red flag

DMC for Safety Only?

DMC needs access to both efficacy and safety to assess risk and benefit

Blinding DMC Members?

Not an FDA or ICH Requirement Imposed by Sponsors to

“Prevent bias” “Avoid over reaction to early trends”

ICH E9

4.1: “Interim analysis requires unblinded access to treatment group assignments”

4.5: “Interim analysis “…involves access to … “unblinded data and results.”

Controlling Bias…Overreaction

The DMC does not make decisions DMC has no vested interest (unlike

company) Monitoring boundaries are in place Degree of empowerment of DMC comes

from Executive Committee and is described in the Charter

DMCs in pre-NDA Setting

Studies in pre-NDA setting seldom stop early for efficacy

Need for adequate safety data (ISS requirements) will often override any early efficacy trends

Monitoring boundaries make it unlikely that effect is significant enough to stop for efficacy

EXAMPLE AMD

Limited phase II dosing information First major entry into patients was two

large Phase III studies. Limited safety data

Efficacy endpoint: mean difference of > 2 lines between treatment and placebo after two years of therapy.

AMD Example

Two large multicenter Phase III studies to show efficacy and establish safety in patients with wet AMD

Primary endpoint: slow vision loss relative to placebo using standard eye charts

Treatment duration: two years Company: Miravant

AMD Example

Monitor for Safety Only AEs, labs, other safety information Interest in acute, untoward decreases in

vision• Examine

Decreases of 3-6 lines Within initial 3 months On individual patient basis

DMC was blinded

AMD Example

FDA mandated post hoc adjustment of p-values since DMC viewed data related to efficacy

AMD Results from 2 Year Follow-Up

Proportion of Patients Losing <2 Lines of Vision___________________________________________________________________

Overall test of no treatment p-value vs.

Study Treatment N n (%) effect placebo_

98EA001 Placebo 64 27 (42.2) 0.0045 SN050 107 62 (57.9) 0.0456 SN075 108 39 (36.1) 0.4293

98EA004 Placebo 55 23 (41.8) 0.1200 SN050 124 72 (58.1) 0.0444 SN075 117 59 (50.4) 0.2908

Summary

Patient safety is key DMC is an independent group Charter describes roles and

responsibilities Monitoring Boundaries are needed DMC needs to be unblinded and needs to

assess both benefit and risk