1. Data Monitoring & Patient Safety Estelle Russek-Cohen CDER OB

2. Disclaimer
The views expressed are those of the author and should not be construed as FDA’s views or policies
Note: I have been a statistician in each of the three medical centers but each center evolves…..conversations with staff in the three centers have been helpful.

3. Center for Biologics Evaluation and Research
Center for Devices and Radiological Health
Center for Drug Evaluation and Research
National Center for Toxicological Research
Center for Food Safety and Applied Nutrition
Center for Veterinary Medicine
Office of Regulatory Affairs
Center for Tobacco Products

4. Data monitoring committees
An external group that has two key roles:
Protecting patients: considering benefits and risks
of an intervention.
Help maintain trial integrity in a clinical trial
With or without a DMC, sponsors of clinical trials are expected to protect patients and maintain trial integrity.

5. When are DMCs required? When cannot obtain individual informed consent
Typically in treatments for trauma subjects
Need community consent
Requirements spelled out in guidance
FDA Guidance for Institutional Review Boards, Clinical investigators and Sponsors Exception from Informed Consent Requirements for Emergency Research (50.24 studies)
IND or IDE required. FDA had an Ethicist on the review team!
NIH funded studies may come with additional requirement for a DMC

6. Role of Guidances at FDA
Current thinking on a topic
Not same as regulation
Draft guidances are posted to solicit public comments
Guidances can be issued by agency, a center, an office, …..
FDA Final Guidance on DMCs:
Guidance from 3 medical centers (CDER, CBER and CDRH)
Some things have changed even if the basics have not.

7. Other guidances and safety reporting
Drugs and biologics (safety under an IND)
2012 guidance…individual serious event reporting
2015 draft guidance….aggregated events and causality
Unanticipated serious events must be reported ASAP if causality is clear
eg acute event right after drug administered (by 7 days or 15 days)
If an aggregated analysis of a serious event suggests causality and the event is not in the Investigator Brochure or events are more common than expected…
a report to FDA is expected
Medical devices (safety under an IDE)
Unanticipated serious events must be reported ASAP ( by 10 days)
Everyone submits annual reports (IND/IDE) and safety summaries come with marketing application (NDA/BLA; PMA/510(k)).

8. DMCs DMCs are expected to be independent of the sponsor
Recommended in FDA Guidance when:
studying patients with severe morbidity and/or mortality
studying vulnerable patients (eg pediatrics, elderly,…)
Primary roles: protect the subjects in the study from unwarranted risks and help maintain trial integrity
Helps to put knowledgeable experienced people on the DMC.
Want people with minimal to no conflicts of interest

9. Since 2006 More trials have a DMC (see clinicaltrials.gov)
Greater use of DMCs for a program rather than a single trial
See more in earlier phases
NIH requires them in many instances
Use of DMCs to implement some adaptive designs
Increase in multi-regional trials and that can add complexity
Charters for DMCs seem to be more complex
FDA (Under IND) is asking for review of aggregate data for safety reporting: assessing causality of serious adverse events

10. Trial Management can be complex: vary with size of trial & safety concerns
Steering committee (often includes sponsor employees)
Internal safety monitoring groups (role vs DMC?)
Endpoints adjudication committee (if appropriate)
Adaptation committee (if adaptive)
IRB: Institutional review board(s): can have centralized IRB structure
Data monitoring committee:
…do not want a DMC adjudicating endpoints
…do not want a DMC designing adaptations after looking at data
Important: knowing when to communicate and how.
who has which responsibility (eg meeting accrual targets..)

11. Expertise on a DMC
Right medical disciplines…ability to add expertise if warranted
Some (especially chair) ought to have experience
Statisticians need to have the right skillset:
Must understand Bayesian or Adaptive aspects in the SAP
Practical experience with data and clinical trials
Often an independent statistician generates reports to the DMC.
If internal statistician prepares reports….what firewalls are in place?
Should have policy on conflict of interest and process to vet members for conflicts of interest

12. DMC charter Criteria for selection of members
Meeting times and who can attend open and closed sessions
Planned analyses in protocol and SAP
Some challenges: is protocol and SAP final by time trial starts?
Maintaining confidentiality
Who will have access to what information (eg indep statistician)
Role in study conduct; interim analyses;
DMC should be informed of FDA concerns on safety

13. Statistical Methods: DMC may do exploratory analyses
Not all methods will be in the SAP submitted to FDA
DMC may need to consider benefit and risk and unblinding may be essential to sensibly reach a conclusion
Since DMC is likely to hold open and closed meetings, it is imperative that any summaries presented in open meeting not unmask results prematurely. (If randomization ratio is not 1:1, think twice about how results are reported. )

14. DMC versus Internal Safety group
Internal group ought to be blinded to/firewalled from efficacy data
If possible safety signal is identified, need to designate how everyone is to be informed. Since timelines are short for mandatory reports to the FDA, it is best if this is spelled out in advance.
Safety signals may need to be communicated in the investigator brochure and identified in informed consent
Any safety signals from internal group ought to be communicated to the DMC
Obviously egregious safety concerns may result in halting the trial.

15. DMCs used in place of internal safety monitoring group for aggregate safety analyses
Is the DMC meeting often enough to spot issues in a timely fashion?
Who will inform the FDA including actions needed (see 2015 guidance):
Stop the trial?
Amend the informed consent?
Amend the investigator brochure?
Note: threshold needed to modify brochure or consent form can differ from putting specific info in the label after drug is approved.
Sponsor needs to have a safety surveillance plan in place and DMC would need to have input.

16. Safety is harder than efficacy
Cannot anticipate all safety concerns; Studies with active controls may pose challenge
Safety coding isn’t always suited for aggregating events for analysis
Statisticians should expect to collaborate with MDs
Acute events such as anaphylaxis close to administration of a drug or vaccine could make causation appear obvious but anaphylaxis isn’t always captured via a medDRA code
If you are treating patients with a serious condition, some events may be due to the underlying medical condition (eg CVOT randomized trials in diabetics are run because diabetics more susceptible to cardiovascular disease)
FDA is not anticipating additional randomized trials for every safety concern, but informing investigators and subjects about potential risks is important.
Multiplicity adjustments for safety at FDA….. Are rare.

17. Role of statistician……
Hypothetical Case of a vaccine for healthy subjects:
Vaccine arm…..2 events out of 5000 subjects
Placebo arm…..0 events out of 5000 subjects
What do you know about the rate of the event in the background population? Suppose it is a rare autoimmune disease that occurs less than 1 in 100,000.
In vaccines with healthy subjects: may know the background rate of some adverse events and expect the physician to pick up on it.
Clinical significance is not statistical significance

18. Statisticians can find patterns:
It will never be one size fits all:
Depends on:
patterns of drug exposure,
if there are multiple dose groups in the trial,
anticipated heterogeneity in patient population
Beware of issues like pooling across previous trials to develop thresholds (eg Simpson’s paradox)
….FDA Draft Guidance on Meta-Analysis for Safety for Drugs and Biologics
If using external data to generate thresholds be aware of data quality, limitations and relevance to current trial
Graphic and Statistical methods can be useful.

19. Conclusions
Statisticians have a key role to play in monitoring a trial including monitoring for safety
However, statisticians and medical experts have to work together for the best result.
Don’t expect a single approach to managing a trial
Don’t expect a single statistical approach to identify signals…
however research in this area is always welcome.