Data Safety Management Boards (DSMBs), also known as Data Monitoring Committees (DMCs), are a complex issue. They are described in two similar documents, one from the FDA from 2006 and the other from the EMA. Another interesting document on these committees has been put out by the UK National Health Service.
This posting and the concluding posting (Part 2) will cover some of the thorny issues that face DMCs now that we are some 8 years post the issuing of the FDA and EMA guidelines on DMCs.
These documents describe “A group of independent experts external to a study assessing the progress, safety data and, if needed, critical efficacy endpoints of a clinical study. In order to do so a DMC may review unblinded study information (on a patient level or treatment group level) during the conduct of the study. Based on its review the DMC provides the sponsor with recommendations regarding study modification, continuation or termination.” (From the EMA). The members usually include physicians (clinicians), a statistician, a safety/PV expert and perhaps an epidemiologist and others as needed. There is also non-voting member who handles logistics, meeting scheduling, minutes etc. A DMC is usually convened in studies that are large, global, multi-center when there are significant perceived risks such as a drug with known or suspected toxicity, patients who are very ill or at risk or vulnerable (aged, babies, immunocompromised, etc…). However, a recent trend is to use DMCs in early phase II and even phase I studies where significant risks are perceived to be present. I am using the words “trial” and “study” synonymously here.
The DMC reviews and approves the protocol and then, on a continual basis, is supplied data from the trial or trials to rapidly identify any safety problems, logistical problems, the continued feasibility of the trial etc. The DMC makes recommendations to the sponsor on whether to continue or stop the trial, make modifications in the trial, alter the consent and/or investigator brochure etc.
In most cases the DMCs function well and efficiently. Their views are taken by the sponsor and put into place (sometimes unhappily).
However, there are some issues and controversies around DMCs which I will discuss here. These represent my own views based on having set up DMCs when I did clinical research and on having served on DMCs as the Drug Safety/PV member.
Blinded vs Unblinded
As someone who has been examining enormous amounts of data over the past 30 years, it is my impression that examining blinded data is far less useful than examining unblinded (unmasked) data. In some cases there is a mix of unblinded and blinded data using different dosages or dosing schedules making comparisons difficult between arms of a study or between studies. In these situations it is often useful to assume that all of the blinded data is the active drug being studied and presume, for the purposes of blinded data analysis, that the safety issues and SAEs are due to the drug. This is, in effect the worse case scenario. It is far easier to examine the unblinded data directly and come to clearer, non-hypothetical conclusions.
Usually when the blinded data suggests an issue, the DMC then requests that the sponsor unblind the data and the DMC then reviews the unblinded data. This can produce unhappiness by the sponsor (particularly if they do not have the staff or software and procedures to easily unblind the data only to the DMC). It can also produce significant delays before the unblinded data is delivered to the DMC thus increasing the time needed to analyze a potentially dangerous safety matter.
In a different context, the FDA and EMA have required that expedited reports (“alert reports”) be sent to the agencies unblinded indicating that blinded data is difficult if not impossible to interpret by the agencies particularly for individual cases. In the case of a DMC, individual case reports and, in my view, aggregate data can also be difficult to interpret.
Can DMCs function well with unblinded data? Yes in some situations, particularly where no safety issues arise! When a safety issue does pop up though, it can be hard to deal with. Imagine one case of Stevens-Johnson or some other SAE that is likely related to the drug. It is important to know which treatment the patient received.
Multiple DMCs
Some companies set up multiple DMCs – one for each trial with no overlapping of members. Thus each DMC sees only the data from the one trial they are overseeing even if other trials with other DMCs are very similar or even identical. This is rather risky in my view. Seeing incomplete data on the active substance gives a limited picture of the true safety profile. This “divide and conquer” situation may mean that safety issues are not picked up. Each DMC might see one very unusual and unexpected SAE and presume it is unrelated to the study drug whereas if they had seen several (say one from each trial) they might think there is indeed a possible causal relationship. As noted, the ability to discern whether a single case report either in trials or after marketing is difficult; the first really unusual and unexpected SAE (e.g. mitral valve disease developing in young women on a drug combo or worsening of NYHA Class III or IV congestive heart failure with immunomodulators) is felt to be unrelated since there is no pharmacologic or clinical reason to suspect a relationship to the drug in question. It is only after the second and third and fourth one that a relationship is suspected.
Thus it is a much better, again in my view, to have a single DMC covering all studies for that drug or, at least, all studies for related indications/patients. The DMC will have a broad overview and will be able to combine data and clinical impressions across multiple studies and patient groups to come to more solid safety conclusions.
This can be very hard and time-consuming. It may require a significant commitment on the part of the DMC members who are usually quite busy. I have been involved in such DMCs and, as the studies progress, we may see thousands of patients’ data and millions or more data points. The CIOMS line listings, tables, charts and analyses may be hundreds or thousands of pages long. There are many ways around this including, what I have seen work very well, is a section by section summary of the enormous documents by the statistician (e.g. “notice that in Table 7.2 there are greater numbers of patients with increases in direct bilirubin in the high and mid dose groups…”).
Using the DMC in the Actual Operations of the Study
This is a situation that is particularly troubling to me. I have seen situations where a DMC is constituted for a study (or several studies) and is asked to make causality judgments for each SAE on behalf of the sponsor. That is, in many jurisdictions (e.g. the US) both the sponsor and the investigator are expected or required to make causality determinations for each SAE. In some situations, often with small companies, the sponsor will not make a judgment of causality using only the investigator’s causality. In other situations, I have seen the company have the DMC make a real time causality judgment on each SAE. This to me is not the role of the DMC. This is the role of the medical monitor in the sponsor or CRO or, for complex issues, an Adjudication Committee. An Adjudication Committee is a group of experts asked to validate a particular SAE or test and make, sometimes, a causality judgment. For example, with a drug that may produce severe arrhythmias, one or more cardiologists may be asked to interpret the electrocardiogram (is this really torsades de pointes?) and is it related to the study drug, comedications, underlying disease etc. This can be done blinded or unblinded.
If the DMC makes this judgment for each case, they are now active participants in the day-to-day operation of the study rather than outside overseers doing a double check on the sponsor/CRO’s handling of the study. The DMC may find it hard to change its mind or give an objective view if they are actively participating in the operations of the study.
Sometimes, the DMC may wordsmith the changes to the IB, consent etc. that they recommend. The DMC should probably not “dictate” the actual wording. Rather they should indicate to the sponsor that the consent should be changed to inform the patients and health care professionals of the need to do a cardiac evaluation before entry into the study. Most DMCs would usually review and approve the changes made by the sponsor but leave the mechanics to the sponsor.
Thus the DMC is the “advisory board and overseer” rather than an actual participant in the study.
Further issues on DMCs will be covered in Part 2.
