FDA Explores New Antiretroviral Trial Designs: Interview with David Feigal, M.D.

On July 14 and 15, at a public meeting of the FDA Antiviral Drugs Advisory Committee, experts from government, industry,academia, and the community will explore new approaches to designing antiretroviral drug trials, based on current knowledge of HIV disease (see announcement in
AIDS TREATMENT NEWS #272, June 6, 1997). On June 12 we interviewed David Feigal, M.D., M.P.H., Director of the Division of Antiviral Drug Products of the U.S. Food and Drug Administration, about current problems in clinical trials, what is needed now, and the FDA's plans for the July meeting.

Background

A new anti-HIV drug can get "accelerated approval" by showing statistical proof of benefit in viral load and other indicators of HIV disease progression. But then the drug must undergo much larger "confirmatory trials," to show statistical proof that it reduces death or major opportunistic infections. These large trials are sometimes called "clinical endpoint" studies, because a participant is recorded as having reached an endpoint in the trial when a major disease event occurs. Because clinical endpoints are relatively rare, and depend on many factors besides the drugs used, these trials usually need to be very large -- usually with more than a thousand volunteers -- to get statistical proof that one of the drug regimens being tested is better than another. These large confirmatory trials have become increasingly problematic as knowledge of HIV disease has advanced, and volunteers do not want to continue using treatments which are not working for them.

The ideas which the FDA is now exploring, which it explained to treatment activists in a community meeting on May 16,would leave accelerated approval as it is today. But companies could choose whether to run the current kind of clinical-endpoint trial, or a new kind of confirmatory trial,which in many cases could record effective viral suppression over a period of time instead of recording clinical endpoints. In these trials, each volunteer would quickly stop any treatment which was not suppressing the virus, be counted then as having reached a treatment failure, and switch to whatever treatment they and their physician decided. This would be very different from the current system, where participants are encouraged to stay on their assigned study regimen until the trial is over, or until they become seriously ill -- which allows viral resistance to develop when participants stay on regimens which do not suppress the virus completely.

By using modern understanding of HIV to individualize the process of dropping ineffective regimens, the new approach should result in more ethical trials, which will be easier to recruit because they are more attractive. Also, these trials should avoid the bias which now occurs when volunteers recognize treatment failure which is not handled in the protocol, and drop out on their own.

Interview with Dr. Feigal

AIDS TREATMENT NEWS: The FDA is considering changes in the requirements for confirmatory trials of antiretrovirals, to not require clinical endpoints in some cases. What is themain problem these changes are intended to solve?

Dr. David Feigal: I believe that the central problem with current clinical trials is not clinical endpoints, but the concern about having suboptimal study arms. For example, a study may compare two treatment regimens: an old one that we know something about, and a new one, to see if the new one is better. People do not want to put the volunteers at risk for losing drugs that could personally benefit them, by administering the drugs in such a way that the study participants are likely to become resistant, or to become intolerant in some other way.

And because many of the drugs are related to each other, there really are not that many of them. Even though eleven antiretrovirals have been approved and there are several others with more limited availability, many have overlapping resistance, or similar toxicities for patients, that limit what you can do with them.

So we have looked at what are the ways that a treatment regimen can be suboptimal. And much has been learned over the last several years, from studies of drugs' effects on viral load. We have become quite convinced that viral load measures are sensitive enough to detect when a drug regimen is effective, and they are also sensitive enough to detect when a regimen has lost its effectiveness, which can happen from a variety of reasons.

Back when there were not many drugs, there were not many choices, other than take one for as long as you could; there was only one drug, or only two that were available. But now that we are individualizing therapy, there is much more in the art of assessing whether a given combination of drugs is producing an adequate response, and trying to get an adequate duration of response. So whether or not the person is someone who is at risk for developing infections or clinical endpoints, we need to make sure that people take the drugs in such a way that they will respond to them as well as possible, for as long as possible, and with the least interference with the use of whatever other drugs are left for that person to take.

ATN: Then the basic idea seems to be to design the protocol so that people stay on the therapy as long as it is working for them, in terms of keeping viral load completely suppressed?

DF: That's right. If you had a drug combination that had a good response in a group, the old approach would be that you take those drugs until you had a clinical event, or until the study ended, or until CD4 counts fell in half, or something else like that. The new approach would say that once someone starts on a study regimen, the first thing we will do is test whether or not that person got an adequate [viral load] response. If not, that is a failure already, and there is no reason for that person to continue to take that particular regimen. And for the participants who get a good response, then of course we want to see how long that response will last.

This is a different approach than we had in the past. It is not the fault of some of the old trials, which did not even have real-time virology [viral load tests which get accurate results without having to be batched and run later] available to them at the time those studies were done. But in the past,many of those participants would not have had a satisfactory response in the first place, and some of them would have lost the response; then they just were followed, while their viruses were developing resistance to the drugs that they were replicating in the face of. So we think we can do the trials in a way that is not only more ethical, but also will provide the kind of information useful in clinical practice.

ATN: You will be looking at something like a 48-week period for the main part of the study?

DF: Yes, one of the major goals of HIV therapy is to be able to have an effect that lasts as long as possible. You cannot determine if something works for a long time with a four-month study. We want the study duration to be long enough to show us how long some of these drugs work. Of course we would like them to work longer than a year; if we had a well-tolerated regimen, it would be great if it worked for multiple years. But a year seems to be a reasonable compromise.

If a company has a new agent that is particularly promising,or has some information based on earlier data, it could seek accelerated approval, and then get the additional data over a longer time period.

During that time, the trial would follow people who had satisfactory responses, and then try to understand why some patients were losing their response. Were they losing it because resistance had developed? Because they had started a new medication for some other reason, and there was drug interaction? Were they losing it because the regimen was too hard to comply with, and they became careless about how faithfully they took the medication? Much of this information we never systematically collected; we had bits and pieces of it. But this is what you need to know to use these drugs in the modern era.

ATN: I understand that in the first part of these trials, you will look for the total viral load drop?

DF: Part of the reason for doing that is that there still is a need to understand how potent the different drugs are. And for that, you need to have complete data on people, at least for a short period of time.

So we hope that one feature of the study design will be a nearly study period, to assess how many patients get an adequate drop -- and that companies will study a wide enough range of viral load to get a good measurement of that.

One question is, what is the best way to characterize thedegree of viral load drop due to a treatment? Should it be the percent of patients who become undetectable by current assays? Or should it be the size of the initial drop invirus? It is not either/or; you want to know both. If we just wanted to know the percent of patients whose viral load became undetectable, that would create an incentive to just study early patients with low viral levels; those are the easiest patients to suppress.

We think we will get a better picture, and a better estimate of the potency, if we say that we want to answer our four big questions:

* How deep is the viral load drop;

* What percentage of patients get a satisfactory response, on an absolute scale [not in comparison with the size of the response in other treatment arms];

* How long does the response last; and

* Why is the response lost, and what are the consequences of losing it?

These are the kinds of questions that we want to answer with the protocols.

ATN: Concerning the July 14-15 meeting of the Antiviral Drugs Advisory Committee, what is the format and purpose of the meeting?

DF: We have asked the companies, academic centers, and government study groups that have data -- data that can answer some of these questions that are needed to design these protocols -- to come in, not to present their whole studies, but to present parts of them. For example, some design questions would be influenced by how much variability there is in the same person when you just measure the virus from day to day. So we invited people who have done studies which generated data on variability, asked them to come in and make a presentation.

The format of the meeting will be for us to break the questions apart. If the question is, for example, determination of treatment loss of response (so-called"treatment failure," which I prefer to call loss of response), how much do people bounce when they are at a low viral load level? How often do you see someone who is at a very low level (of virus) have a little spike, but nothing else happens over time? Many of these trials were not trying to individualize therapy, so they have observed what happens over time; they will be able to help us design the studies so that one does not yank a therapy too quickly, because of background noise, and give us a better sense of how the assays perform.

If you step back and look at the big picture of what we are asking the Committee, we are asking a couple of very fundamental questions. One, is there adequate evidence that there is clinical benefit from lowered viral load per se? If reducing viral load really is a goal of therapy, then that should be the indication of the therapy. Now the drug labeling sort of says, "Indicated to treat HIV." That does not really tell you exactly what you are doing and how you are doing it.

We are asking, for antiviral drugs, have we moved to the era where we should focus in on the virus itself?

The other change, which is a little more subtle, is that in the past, when we looked at surrogate markers like the CD4 count, or viral load drop, the emphasis was on getting complete data on everyone for a fixed time period, with everyone trying to stay on drug. We wanted to see which treatment had the better group average response. Obviously one of the problems with those studies was that the patients whose markers were doing poorly left the study, and artificially affected the measure -- making the counts look better than they were, because people with unfavorable counts would drop out.

Now, instead of asking what is the average effect of taking the regimen for a period of time, we are asking what percentage of people get an adequate response, and how long does that last? It is much more of a time-to-event. You only stay on a therapy if it works; and you only keep taking it if it continues to work. We are essentially asking the Committee, is the data on viral load good enough that we can rely on it to do this?

Although the issue is phrased as whether we are leaving clinical endpoints, from our standpoint we are not. We hope that companies will continue to study new products in a spectrum of HIV conditions -- including patients with very low CD4 counts and high viral load, that are at high risk for developing infections and other complications. In those trials, they should be able to tell whether or not totally suppressing the virus with one regimen is just as good as totally suppressing it with another. Or is there some other problem that emerges?

Similarly, there should be studies of patients with early disease, to try to get at the questions you want to answer about early disease.

To us, the issue is not are we going to stop studying patients that are developing infections. The real issue is to see how well we can combine drugs, to individualize therapy, and to see how useful a product is within a combination, indifferent stages of the disease.

ATN: In this meeting, will there be any focus on the problems of getting the companies to work together, and be willing to have their drugs tested with competitors' drugs?

DF: I am sure that this issue will be raised in the public comment section. I know that has been an important issue for the community, and for practitioners who want to know information about certain options. It is clear that there are times when decisions about what types of studies to do are made by companies based on their interests in seeing what they can do with their own products; as you know, several of the companies have multiple products. They may well be missing opportunities to find combinations which are particularly effective, which would require testing their drug along with their competitor's product. I hope that some of the independent studies that are often done by groups such as the ACTG will level that playing field a little bit.

Given the way that the drug development structure is based often on economic incentives, a company would have to see that if they found that their product was useful with their competitor's product, that would have some economic advantage to them. Their first take may well be that this would not be as advantageous as selling two of their own products, so they will test that first.

But there is much screening that needs to be done, and many products to be tested. Hopefully, by basing clinical trials more on viral load and less on studies that look for disease progression, the studies will be small enough that we will be able to more efficiently screen promising combinations. And studies that in the past might only have been feasible with large-scale corporate or government sponsorship could then bedone on a more modest scale, to identify hot leads to follow.

ATN: Is there a way for the public to submit written comments to be considered by the Committee, if one is not going to be there in person?

DF: Send them to Rhonda Stover, the executive secretary. They can be provided as background to the Committee members, either in the mailing we send them before they arrive, or in their packet at the time of the meeting.

[Mail or fax comments to: Rhonda Stover or John Schupp,Center for Drug Evaluation and Research (HFD-21), Food and Drug Administration, 5600 Fishers Lane, Rockville, Maryland 20857, phone 301/443-5455, fax 301/443-0699. Those who would like to speak during the public comment period at the meeting, July 14 from 11 a.m. to noon, should notify them by July 7. The meeting will be 8:30 a.m. to 5:00 p.m. each day, in Room 204 of the Armory Place, 925 Wayne Avenue, Silver Spring, Maryland, walking distance from the Silver Spring Metro stop.]