Clinical Trials and Viral Load: Statement for FDA Community Meeting May 16

Introduction

We wrote the following statement to suggest issues for discussion at the FDA community forum on May 16, concerning using viral load instead of "clinical end points" in certain trials of anti-HIV drugs (see announcement in AIDS TREATMENT NEWS #270. This introductory section provides background for readers who are not familiar with current debates and issues in clinical trial design.

In the early years of the epidemic, due to historical accident, clinical trials of AIDS drugs developed from the model of cancer trials. The traditional "gold standard" in cancer trials was five-year survival. Since five-year trials would clearly slow drug approval, there have been years of debate about whether substitute ("surrogate") cancer measures, such as tumor size, were an acceptable substitute for survival for drug approval.

In AIDS, the cancer standard of survival was quickly relaxed a little, and the traditional gold standard for AIDS drugs became improved survival or reduction of major disease progression (usually taken as the development of a new AIDS-defining opportunistic condition) for some unspecified period of time. These are the "clinical endpoint" trials referred to in the article below. In a clinical-endpoint trial, a new drug must prove that it is at least as good as some standard treatment in reducing death or major disease progression in some group of patients.

The first problem with clinical endpoint trials is that they commonly require over a thousand volunteers and take well over a year of actual time on treatment (plus more years to plan, set up, and recruit) to get statistical proof. And more recently, especially within the last year, a new problem has arisen. The importance of lowering viral load (especially to levels so low that it is undetectable) has become widely accepted. Since viral-load trials can be run much more rapidly than clinical-endpoint trials (and can usually get statistical proof with about a tenth as many volunteers as required for clinical-endpoint trials), it can quickly be learned that some of the treatment arms in clinical-endpoint trials are inferior to others in lowering viral load. It is now widely considered unethical to ask volunteers to stay on these virologically inferior treatments for years (see discussions in SCIENCE, April 25).

For many years there has been debate on whether anti-HIV drugs must go through clinical endpoint trials in order to be approved, or whether improvement in "surrogate markers"(especially viral load) would be enough. Several years ago the FDA started a system called "accelerated approval," which allows a drug to be approved based on surrogate markers(without having to wait years for clinical endpoint trials), provided that clinical endpoint trials are run later (to confirm that the surrogate-marker benefit translated into a real clinical benefit for patients). Today these confirmatory trials are creating intolerable problems, mainly because it is no longer considered ethical to keep patients on virologically inferior treatments, now that the importance of viral load has become much more accepted than it used to be.

During the last year and a half, the viral load measurement which has become widely regarded as most important is the proportion of volunteers whose viral replication can be essentially shut off (as shown by undetectable viral load in the blood plasma) for as long as possible. Other measurements, like the group average fall of viral load, are not considered as relevant. Our article below explores the surprisingly strong advantages of using the ability to maintain undetectable viral load as the primary way to judge an antiretroviral drug regimen in confirmatory trials.

--------

Statement to May 16 FDA Community Meeting:

Clinical Trials and Viral Load: Complete Viral Suppression As Primary Endpoint for Confirmatory Trials

The ethical, practical, and scientific problems with the current regulatory requirements for confirmatory trials of antiretrovirals are well known. The biggest problem concerns trials with suboptimal treatment arms. These often try to show clinical-endpoint superiority of a regimen containing the drug being tested, compared to some other regimen which the trial designers or sponsors thought they could get away with -- but which must prove inferior through death or major illness for the trial to be a success. The need now is to articulate consensus and/or develop regulations to encourage workable trials which are ethical in the context of modern medical knowledge, yet provide the answers physicians need and can use.

It is now becoming widely accepted that instead of clinical endpoints, some degree of viral load suppression, for some period of time, should be required for final approval. (Formulating this agreement and providing specifics is the apparent mission of the mid-July meeting of the Antiviral Drug Products Advisory Committee.) Since the time required for showing durable viral suppression is likely to be a year or more, approval should not be delayed until completion of these trials. Therefore, the current system of accelerated approval, which has been successful, should be kept in place.

Complete HIV Suppression, and Trial Design: A Fortunate Synergy

Today it is widely agreed that antiretroviral therapy should aim for complete suppression of HIV replication (to undetectable levels of plasma viremia), sustained for as long as possible. Far from making it impossible to test new drugs, as some have feared, this modern view makes possible a new generation of very attractive confirmatory trials. The key is to use loss of complete viral suppression as the primary endpoint to define failure of the assigned treatment regimen for that individual -- that is, as long as a treatment completely suppresses the virus, it is regarded as successful, but when complete suppression stops, it is counted as having failed. In that case, the individual is immediately unblinded and offered new treatment options (which should usually include new randomizations, for further research). Drug failure due to toxicity or intolerance of the regimen would also lead to unblinding and new treatment options.

Especially near the beginning of treatment, and later as required, frequent viral load testing would be used to detect failing regimens quickly, and volunteers would be offered other options without delay.

"Treatment failure" would not be declared from a single viral load result, if only because of the possibility of testing errors. Exact criteria for virological failure -- either loss of complete suppression after it had been achieved, or inability to achieve it in the first place -- would be defined in the protocol.

Some advantages of this approach:

* Each trial volunteer always receives virologically optimal treatment (or is given the option of trying another therapy). No one need stay in any arm unless the therapy is suppressing viral replication to the greatest extent that can be measured. This attractiveness to patients and treating physicians should greatly help in recruiting.

* These randomized controlled trials would provide the data which is widely regarded today as most central: How well does a regimen prevent drug failure (due either to viral breakthrough, or to drug toxicities)in the patient population being tested?

* Failure of complete viral suppression is already in common use in drug trials, and is well regarded as a virological measure. (It has not yet been used as the primary endpoint for confirmatory trials, because the FDA has required clinical endpoints so far.)

* This trial design avoids most of the need to hide information. Volunteers are of course given their viral load results immediately (since the viral load must remain undetectable, or the volunteer will be unblinded and offered different treatment).

* There is much less need than with previous trials to conceal overall results as they develop, in order to avoid biasing the trial. This is because, regardless of the overall statistics of drug failure, each volunteer is either on a treatment which, virologically at least, is working perfectly for them (as far as can be measured), or is already changing to a different treatment. Participants are unlikely to abandon a (blinded or unblinded) regimen which works this well for them, in favor of a treatment which may not work, based on statistical results from other people.

* Long-term data on treatment success or failure are critically needed -- and this kind of trial can obtain such data much more easily than conventional designs. Trials could regularly be planned to run for five years or more. This is because participants are not kept on a treatment which is not working for them. Instead, their failure and abandonment of the treatment provide the data for measuring the effectiveness of the antiviral regimen. Therefore, the only volunteers who will be kept on a regimen for years are those for whom it is working very well. They will be unlikely to want to change.

Also encouraging long-term continuation of randomized controlled testing is the fact that results can be reported as they develop, with little fear of subsequent bias. There is no need to close the trial or risk its future integrity so that results can be published for general use.

For the same reasons, approval and marketing of the drugs should have little effect on these trials -- further encouraging the generation of long-term data from randomized treatment, regardless of when accelerated approval and final approval occur. The key difference from most current trial designs is that the only people on the long-term treatments are those who are doing very well, who therefore have no incentive to leave the study -- while those who left after treatment failure are not dropouts responsible for missing data, but instead reached a primary endpoint exactly as planned in the protocol.

And sponsors will be happy to pay for long-term follow up for as many volunteers as possible, because that will mean that their drug is successful. The more people who stay on the treatment long term, the more valuable the drug is. This creates a strong economic incentive to collect long-term data.

* This kind of trial might reduce the bias which now occurs because volunteers who are more ill are less likely to report for study visits. Those who are doing poorly would probably already be recorded as treatment failures, so any appointments they miss will not affect the primary endpoint of the trial.

Followup after treatment failure -- even after leaving the protocol -- is also important. Especially those who have left the protocol need reimbursement for their time and expenses in study visits, since the drugs no longer serve as incentive. It is a great mistake to just drop people after they leave a protocol -- yet that is commonly what happens.

* Because virologic treatment failure (loss of complete suppression) is likely to occur much more rapidly than clinical endpoints -- with today's treatments at least --these trials should require far fewer volunteers than clinical-endpoint confirmatory trials.

* When treatments improve enough that failure (either from viral breakthrough or drug toxicity) becomes rare, then these trials will keep more of their participants on their original randomized arms for a long time -- effectively becoming CLINICAL ENDPOINT trials, with increasing power as more volunteers remain on their assigned treatment for longer. This provides an ethical way to look for any unexpected disease progression in spite of complete viral suppression. In the most likely case -- that no such disconnect is found-- nothing is lost, since all of the volunteers will have received optimal treatment, all groups will have done well clinically, and the trial will have provided valuable assurances about the drugs.

Regulatory Oversight Needed

Several areas will need continuing FDA and public attention to ensure that confirmatory trials with virological endpoints best serve the public interest:

* What happens after treatment failure? Are volunteers just dropped because the sponsoring company has no further interest in them? Those who have participated in a trial should have assistance in finding a treatment strategy which works for them -- especially since the public could benefit greatly by knowledge obtained in this process.

The medical community needs to know what options are useful after certain drugs have failed -- both to treat those individuals successfully, and to learn how initial choice of therapy may limit future choices. In some trials, subsequent randomization options after failure of the initial antiretroviral regimen will be appropriate for obtaining this information. For various reasons, including the fact that drugs from different manufacturers are likely to be involved, a private sponsor may not want to do this research. It should be public policy to ensure that such studies are done as part of a drug's approval.

* At the very least, long-term observational followup should be part of most if not all antiretroviral trials. Safety and quality of life information is needed for all who are randomized to a treatment, even after they discontinue it.

* Antiretroviral drug development must include testing in advanced or heavily pretreated patients. Any new antiretroviral will be widely used by them. It is unconscionable to collect no data and test the drugs only where they are likely to look best.

* Physicians and patients need other data from the approval process -- data often neglected today. Small, rapid pharmaco kinetic trials should be required, to establish:

- Doses for infants and children;

- Possible dose adjustments for sex, race, or body weight;

- More interaction data with other medications which patients being treated with the drug being tested are likely to use --including illegal or recreational drugs -- when there is any reason to suspect an interaction;

- Per-patient variation in blood levels due to individual differences in absorption or metabolism -- and the effect of such variation on treatment failure.

Ending the previous generation of clinical-endpoint confirmatory trials will save enormous resources which are now mostly wasted. The FDA, supported by professional consensus, should require pharmaceutical companies to use some of the resulting savings to provide a rational package of information to the patients and physicians who will rely on their products.