Skip to content

“Don’t Mess with the DSMB”

July 30, 2010

In the July 29 issue of The New England Journal of Medicine, Drazen and Wood present an editorial of this title:

Since the DSMB (data and safety monitoring board) is charged with ensuring that clinical equipoise is maintained as trial data are accrued, it is considered very bad, even self-destructive, behavior for people who are involved with the study to interact with DSMB members on trial-related issues. Traditionally, there has been a wall between investigators, sponsors, and the DSMB. This wall prevents preliminary findings from leaking out in ways that would prejudice the trial. For example, if it was known that the DSMB was examining a marginal increase in cardiovascular risk in a trial, then trial investigators might bias future recruitment by excluding patients at risk for such events. In the proper performance of clinical trials, you “don’t mess around with the DSMB.”

For people unfamiliar with clinical trials lingo, this paragraph and article read like a foreign language, but the point is really important. To understand what the DSMB is and why you want one, we need to start with the design of a clinical research project.

Study Design: Predicting the unknown

So you want to see if Drug A lowers blood sugar in patients with diabetes. You predict a modest effect, and you even run a small trial. Based on those data, you design a large study. Clinical trials of this sort often require large numbers of patients to show a difference between Drug A and no drug with a reasonable certainty that the difference is not merely due to chance. Of course, you must also consider that your drug does not work. You must include enough patients in your study that a lack of effect is also unlikely to be due to chance, a concept referred to as the power of a study. Sample size and power analysis on preliminary data estimate the number of patients necessary to meet these criteria. Some patients won’t agree to participate in research, while others will have some reason they should not be subjects. Factoring in these probabilities leads to a number of patients that must be screened to get the correct number of patients in a study. Often, each study condition requires several hundred patients.

One doctor, one medical center, even one city is unlikely to produce enough patients during a reasonable time period to test a drug. Groups of investigators (the doctors running the study) organize multiple centers, often across national borders, to achieve enough patients to show differences.  Each center may only recruit a handful of patients to any given trial.

The gold standard for drug trials are “double-blind” studies. In these trials, neither the patient nor the physician observing the patient knows what treatment is being used. While this condition removes potential bias from the trial, it also means that treating physicians cannot note problems associated with a treatment or lack of treatment.

The Data Safety Monitoring Board

Say the effects of Drug A are far more dramatic than you guessed. What if a beneficial effect can be shown after only half the patients have enrolled in the study? It would not be ethical to continue to withhold a beneficial treatment from the subjects getting placebo.

Say Drug A has effects as you estimated, but it also causes side effects. If those side effects are serious, continuing the study may no longer be worth the risk to your subjects.

What if Drug A has no effect? You may have sufficient power to prove its futility before enrollment is complete. While continuing the study may not harm your subjects, it will use up valuable resources you could use to develop another treatment.

While working under double-blind conditions, often with only a tiny fraction of the subjects at any given recruitment center, it is impossible to identify any of these issues that would lead to early trial termination. If the investigators become worried about something, they would have to break their code to look at the data early. They would no longer be masked to patient treatments, and the trial could not continue.

Data Safety Monitoring Boards (DSMBs) allow the data to be monitored while keeping investigators masked. The DSMB must be independent of the study investigators. It must include a statistician, as well as physicians familiar with the condition under treatment and any potential side-effects of the treatment. At defined intervals, the DSMB examines subject accrual and data. The board must decide if the potential benefits of the treatment still appear to outweigh its risks; if the study is on-track for enrollment; and if continued study is futile.

Messing with the DSMB

So what went wrong in the studies featured in the editorial? Pharmaceutical companies became aware of potential risks of trial drugs. Instead of requesting that the DSMB examine the data and determine whether or not the trial should continue, sponsoring companies unblinded studies. One of these events involved rosiglitazone (Avandia; GlaxoSmithKline), and the manipulation of the DSMB did not become public until senate hearings earlier this year. Given the ongoing controversy over this hypoglycemic agent, I really wish these data were not tainted this way. Would the DSMB have halted the study anyway? Perhaps; but once the sponsor and investigators were no longer blinded to results, the study could not continue.

Levels of Safety

There are several levels of safety in clinical research. Researchers must convince the Internal Review Board (IRB) that the potential benefits of their studies outweigh the risks. The IRB must also approve the informed consent procedure and any advertising to recruit patients. Long, structured documents outline known potential risks of any study, including loss of privacy. This form must be approved by the Board, and the investigator must follow the consent procedure to the letter. Informed consent allows the subject to decide if they wish to endure the study; only the patient can decide if they wish to take a risk of any sort.

Every year, interim reports must be filed with the IRB to document progress in recruitment. For large or risky trials, the IRB requires a DSMB which will make a separate report to the IRB. The DSMB is the last layer of safety for patients participating in clinical research. Its importance for large multicenter double-blind studies cannot be overstated.

As Drazen and Wood note:

You don’t tug on Superman’s cape

You don’t spit into the wind

You don’t pull the mask off that old Lone Ranger

And you don’t mess around with the DSMB.

One Comment

Trackbacks & Pingbacks

  1. Is Volume Toxic? Part 2 | Whizbang

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: