TORONTO—Researchers in epilepsy are looking to a new type of control group as a solution to problems incurred when the availability of several effective treatments represents—at least as far as clinical trial design is concerned—too much of a good thing.

The solution in question is to use what is referred to as a historical control—or virtual placebo—in place of a placebo arm. Thanks to a spate of dose-controlled comparison trials yielding the right combination of results, the field of epilepsy may offer nearly ideal circumstances in which to test this approach, said Jacqueline A. French, MD, Professor of Neurology at the Hospital of the University of Pennsylvania and Codirector of the Penn Epilepsy Center, both in Philadelphia. “Unlike some of the other diseases … epilepsy actually is something for which there are many treatments,” she said at the 129th Annual Meeting of the American Neurological Association. “That is very, very good; but as far as trials are concerned, that can make it very, very difficult.”

THE SEARCH FOR AN OPTIMAL CONTROL GROUP

Placebo-controlled trials are a time-honored, scientifically valid method of evaluating new drugs for monotherapy, Dr. French remarked —but when effective treatments exist for a given disease, use of placebo to study a new drug raises ethical issues.

With the availability of several established and more recently approved treatments for epilepsy, clinical investigators have turned to adjunctive therapy trials to evaluate new antiepileptic drugs. Although these studies can provide definitive results in the initial assessment of new drugs as add-on therapy for specific seizure types or epilepsy syndromes, adjunctive trials are less useful for evaluating the potential benefits of a new agent and can be complicated by the pharmacokinetic and pharmacodynamic interactions of the dual- or multiple-drug regimens under study.

Because clinicians tend to treat newly diagnosed epilepsy patients with a single drug, use of new antiepileptic drugs in this patient group would likely take the form of monotherapy as well. Yet, although antiepileptic drugs approved for adjunctive therapy can be used as a single agent, some clinicians will not prescribe them without an FDA- approved indication for monotherapy, Dr. French noted.

Thus, the need for monotherapy trials to evaluate new antiepileptic drugs has led to a search for an optimal control group, she said. In several studies, use of an active control—by which the study drug is compared to an established agent—has raised concerns that neither arm produced valid effects, given the specific study population and design chosen for the trial. A more controversial strategy is the use of “pseudoplacebo,” whereby one arm is randomized to a reduced dose of the study medication—a regimen efficacious enough to prevent severe seizures, yet with enough of a difference to allow some seizures to occur and enable investigators to determine whether the study dose is effective.

It was the use of this pseudoplacebo in nine clinical trials in which patients were withdrawn to monotherapy that led, fortuitously, to the advent of the historical control as is now being examined in epilepsy. These were trials involving patients whose epilepsy had been shown to be refractory to previous treatments. In each study, investigators thought that the low-dose control regimen would be safe but would allow a difference to be demonstrated between the two treatments. However, in retrospect, the low-dose regimen was not as safe as investigators originally thought, according to Dr. French.

“As their medications were being withdrawn down to almost nothing, [the patients in these arms] were having a significant worsening, they were having status epilepticus, they were having convulsions that they didn’t have before they went into the trial,” she said in an interview with Neurology Reviews, recounting some of the outcomes. “Therefore, we would say, at this point, that those trials—on an ethical basis—should not be conducted anymore.”

A LEARNING EXPERIENCE

Because these particular studies have already been done, the question then becomes what can be learned from them, Dr. French posited. On closer examination, what she and her colleagues found was remarkable. “The group that was withdrawn to almost nothing behaved in exactly the same way every time a trial was done,” she said. In other words, based on the preestablished exit criteria, which were standardized and remained the same from trial to trial, the percentage of patients in the low-dose control arm who exited—that is, who were withdrawn to monotherapy or otherwise removed from the arm due to the ineffectiveness of the treatment—was between 70% and 100%.

Such similar outcomes could not have been expected, because the pseudoplacebo arms and the patient populations in each of the trials and the way the trials were conducted were all at least a little different. “So you could have found that those arms were all over the map,” Dr. French said. “But amazingly, they weren’t; they, in fact, behaved extremely similarly.”

Because all these arms behaved so similarly, they represent—in the view of Dr. French and her colleagues—a valid historical control. “You know what’s going to happen to people when you take them off their drugs,” she said, explaining the rationale for using a historical control in place of another type of control arm. “Now all you have to do is prove that when you put them on a drug, they behave differently, and you have to prove that by a confidence interval.

“You can now say people withdrawn from an antiepileptic drug have between a 70% and 90% chance of exiting, based on these criteria, with a confidence interval from X to Y, and people who are on an effective drug therefore shouldn’t fall anywhere near those bounds—they should do much better.”

HISTORICAL CONFIDENCE

Use of a virtual placebo displaces the need for an actual placebo because the confidence interval allows investigators to put subjects on active treatment only. “The placebo arm is essentially hypothetical, based on the historical control,” Dr. French said. “Now you can say, ‘I have a drug and only 50% exited or only 30% exited; then, obviously, that’s an improvement over what I think placebo would have done.’ And that’s what the historical control tells you.”

This approach would still allow investigators to use more than one arm in a trial—for example, if they sought to compare two different doses of the study drug or even to compare a new drug with an established drug. One could take several different approaches to randomizing patients without the use of a placebo arm, Dr. French emphasized. For her, the bottom line for clinical investigators—whether they are working in epilepsy, multiple sclerosis (MS), Parkinson’s disease, or any other neurologic disease that has the potential to cause harm if patients go untreated—is clear: Once effective treatments exist, active placebo is no longer an option.

“And by that time it’s too late to set up a historical control, because then you can’t go back and say, ‘OK, now I’m going to do my placebo-controlled trials in order to set up the historical control,’” Dr. French remarked. “So you have to start thinking about this as you’re doing the studies with what you hope will be effective treatments. And the only reason that our historical control has any hope of working is because all of the studies were highly standardized.”

Do the circumstances of those studies make epilepsy unique among neurologic disorders with respect to historical control? Probably, said Dr. French, since she is unaware of it being used in trials in other fields. She understands from colleagues that researchers in MS, which has a large clinical trial database to draw on, are looking at the effects of placebo on patients in control arms. “That’s not quite the same, but they are making efforts to look at placebo activity,” she said. “So they’re thinking along the same lines.”

A VALID—BUT AS YET UNACCEPTED—CONTROL

Although historical control is explicitly cited in the regulations of the FDA as a valid control group, the agency has yet to embrace this virtual approach as a strategy to be widely promulgated for clinical trial design. Dr. French and her colleagues are finalizing a paper validating the use of historical control in epilepsy clinical trials, and they plan to first submit it to the FDA for comment and approval.

“In the end, this all really comes down to not only our satisfaction but the satisfaction of the FDA,” she said. “If they accept [the paper’s findings], that will be the first time that a historical control has been accepted by the FDA as a basis for trials.”

NR

—Fred Balzac

Suggested Reading
Yen A, Simpson E, Haverkamp L, Appel S. Design using historical controls. Amyotroph Lateral Scler Other Motor Neuron Disord. 2004;(suppl 1):61-63.

Return to table of contents