Data and Safety Monitoring

Data and Safety Monitoring

J. Allen McCutchan

INTRODUCTION

Institutional review boards (IRBs) are tasked with the "continuing review" of research (i.e., monitoring studies after their initial approval) but face a multitude of problems in trying to do so. This chapter attempts to clarify the responsibility of investigators, sponsors, and IRBs for monitoring the conduct of studies after their initial approval. It advocates a central role for data and safety monitoring boards (DSMBs) in the monitoring of complex or potentially risky studies.

DataandSafetyMonitoringBoards.

A DSMB is a multidisciplinary group that is usually composed of three to six experts in at least two areas: (1) medical issues (the disease, drug, device, procedure, or outcome measures) and (2) method issues (clinical trials design, data management, and statistical analysis). For some studies, expertise in research and biomedical ethics is also required. DSMBs can monitor the timeliness of accrual, the quality of data collection and management, and the accumulating outcomes to assurethe safety of participants and the scientific integrity of the study.

The terms DSMB and independent DSMB are used interchangeably because a criterion for serving on a DSMB is a lack of ties, affiliations, or interests that might create bias when evaluating data or making recommendations. The independence of the DSMB requires that its members have no professional or financial interest in the outcome of the studies it monitors. On the other hand, the sponsors or investigators convene and support the DSMB, selecting and compensating its members for their services. Moreover, the data that the DSMB reviews are usually processed and presented by data managers who are employees or contractors of the sponsors or investigators. Thus, the DSMB depends on the professional integrity of the data managers for the accuracy of data it reviews.

When Is a DSMB Needed?

IRBs are currently without meaningful guidance on when to require a DSMB as opposed to other models of data oversight. A definitive evaluation of this issue is beyond the scope of this chapter. In general, the factors that suggest that a DSMB is the most appropriate way to monitor data include these:

1. A large study population. Many persons may be harmed before problems are recognized.

2. Multiple study sites. When no investigator treats more than a fraction of the participants, it is difficult to recognize a pattern of increased or unusual problems.

3.Highly toxic therapies or dangerous procedures. This is a critical determinant of how frequently and intensely participants

should be monitored for adverse effects

4. High expected rates of morbidity or mortality in the study population. Problems related to the natural history of disease or aging may obscure adverse events caused by the intervention.

5.High chance of early termination. There is a reasonable likelihood that the study may be terminated early for reasons of safety, futility, or efficacy.

What Does a DSMB Do?

DSMBs usually review data at predetermined intervals during the study, sometimes operating under protocol-specified rules for considering whether or not the study should continue. The DSMB may specify the type and form of data presented, examine the blinded data by group (without being told which group is which), and/or completely unblind the study. Their decisions are rendered as suggestions to the sponsors or investigators and are subject to review and negotiation before their implementation.

Meetings of a DSMB are often divided into an open session attended by the investigators and sponsors and a closed session limited to the members of the DSMB. During the open period, issues of data qualitymay be discussed. In addition, the rates of accrual of participants, outcomes (such as deaths, relapses, or responses), and adverse events may be reviewed for the study as a whole. In a closed session, the DSMB may see the rates of these events in each study arm and discuss the issues of modifying or stopping the study.

DSMBs may conclude that the study should be stopped for one of several reasons.

1. Efficacy-With high certainty, the study question has been answered (e.g., one group is doing better than the others, or no group is likely to do better than any other).

2. Futility-The study question will not be answered when the study is completed, for example, because too many of the patients have been lost to follow-up or stopped the intervention.

3. Safety-Risks to patients are too high.

Most investigators and sponsors immediately accept and implement the conclusions of the DSMB, but occasional disagreements require negotiation. The IRB should insist that it be notified immediately of the DSMB's recommendationsto modify or stop a study.

DSMBs and the IRB

DSMBs can monitor the safety of study participants more rigorously and more easily than IRBs and should routinely communicate their findings and recommendations to all of the IRBs involved in multisite studies. The frequency and content of the reports will vary depending on the safety issues and specific monitoring plan proposed for each study. IRB members and administrators have persuasively advocated this approach in several recent articles.''' Moreover, the National Institutes of Health (NIH) has adopted this policy.'

DSMBs can assist IRBs in their oversight of studies because DSMBs have greater access to the accumulating data than IRBs and can unblind the treatment assignments of either individuals or groups of participants. Because unblinding can be independent of the investigators or sponsors of the study, safety issues may be addressed without compromising the objectivity of the investigators. Knowing the group assignment of individuals in blinded treatment studies is often invaluable for assessing an individual adverse event. DSMBs can make informed decisions that the IRB could not attempt. For example, review of unblinded data by a DSMB can detect excess morbidity or mortality in one arm of a large trial that is not apparent to the IRB or investigators. As discussed earlier, DSMBs may also terminate a study earlier than scheduled because the study has already reached a clear outcome or is very unlikely to do so even if completed. This unique insight of the DSMB contributes to safety because it minimizes the exposure of patients to more risky or less effective therapies.

DSMBs have become standard in the monitoring of most NIH-sponsored and many industry-sponsored Phase III clinical trials over the past decade. Large, multisite clinical trials groups such as the AIDS Clinical Trials Group of the National Institute of Allergy and Infectious Diseases routinely review their interventional studies.' Guidelines for monitoring of Phases I, II, and III studies are available on the Webs'' Despite this widespread appreciation of their usefulness by the NIH, their DSMBs do not routinely communicate with local IRBs, and many IRBs are unaware of their existence.

How Can DSMBs Be Integrated into IRB Procedures? The time to deal with the DSMB issue is at the time of initial protocol review. This is the subject of other chapters in this book. Initially, IRB members and administrators

may be unfamiliar with the functions of DSMBs and their advantages for investigators and the IRB. Data-monitoring plans, including consideration of the need for a DSMB, should be a routine part of evaluating the data management and statistical section of protocols. This may require that the IRB sponsor training, for both investigators and IRB members, addressing the need for DSMBs, and that reminders be added to the instructions for protocol submission and review.

Negotiations with the sponsors of studies to establish a DSMB, ask for reports on their reviews, or request unblinded reviews of serious adverse events (SAEs) are usually conducted via the principal investigator (PI) of the study at the IRB's institution. The PI may not be able to convey the rationale for the IRBs request to the sponsor without help. IRBs should provide explicit reasons for requesting establishment of or action by a DSMB in writing to the PI. These requests may explain (1) the features of the study that require establishment of a DSMB, (2) the specific reports required from the DSMB, (3) the reason that SAEs need to be evaluated by the DSMB, or (4) additional information required by the IRB, such as "stopping rules" for the study.

For example, the letter to an investigator might contain remarks like the following:

The plan and resources for data collection and monitoring were not adequately outlined in your submission. Please indicate the personnel, site, computer resources, and monitoring procedures you will employ as required in the data and monitoring section of the outline for submission of protocols to the IRB. Your submission indicates that (DRUG X) has been associated with (SEVERE TOXICITY) in a substantial proportion of patients in prior studies and that you intend to treat patients at a more advanced stage of disease than in prior studies. For these reasons, the board suggests that you convene a DSMB to review the safety of the study. Specifically, after treating the first 50 patients and' after each group of 50 patients thereafter, the DSMB should assess and compare the accumulating rates of (SEVERE TOXICITY) in the two arms of your study and communicate to the IRB that continuation is appropriate. Also indicate and justify in the protocol what rate of toxicity in the DRUG X arm would trigger the DSMB to stop enrollment and reevaluate the safety of the study.

Approval or continuation of the protocol should depend on an adequate response to the IRB's requests for better definition of the monitoring plan. Investigators at cautious institutions may be unable to participate in the study if the sponsor or DSMB is unwilling to respond to such requests. Because individual IRBs require that DSMBs be convened and communicate regularly with IRBs, they may be perceived initially as "difficult" by study sponsors or local investigators. This may appear to place local investigators in an awkward position, especially if they have no control over the content of the study protocol. However, the protections afforded to the research subjects, the investigators, and theinstitution greatly outweigh this disadvantage. Moreover, as multiple IRBs make similar requests, investigators and sponsors will understand that, for many studies, regular monitoring by a DSMB is the standard of practice.

Conclusion

Data and safety monitoring is one of the most complex and important aspects of the process of protecting human research subjects. The role of the IRB in the data safety monitoring process is discussed in detail in several other chapters in this book. This chapter has focused on the most rigorous approach to ongoing data monitoring, the data and safety monitoring board. For IRBs to function optimally from the standpoint of protecting research participants in compliance with federal regulations, it is important for the IRB to understand when it is appropriate to require a DSMB as part of the initial research plan and how the IRB and DSMB should interact during the life of the study.

References

1.Gordon VM, Sugarman J, Kass N. Toward a more comprehensive approach to protecting human subjects: The interface of data safety monitoring boards and institutional review boards in randomized clinical trials. IRB: A Review of Human Subjects Research 20(1):1-5, 1998.

2. Bankert E, Amdur RJ. The IRB is not a data and safety monitoring committee. IRB: A Review of Human Subjects Research 22(6):9-10, 2000.

3. National Institutes of Health. Guidance on reporting adverse events to institutional review boards for NIH-supported multicenter clinical trials (from NIH Guide for June 11, 1999). Access date 29 March, 2001.

4. Ellenberg SS, Myers MW Blackwelder WC, HothDE The use of external monitoring committees in clinical trials of the National Institute of Allergy and Infectious Diseases. Stat Med 12(5-6):461-467, 1993.

5. National Institutes of Health. NIH policy for data and safety monitoring, 10 June, 1998. Access date 29 March, 2001.

6. National Institutes of Health. Further guidance on a data and safety monitoring board for Phase I and Phase II trials, 5 June, 2000. Access date 29 March, 2001.