The Cycles of History: With Good Intenntions, Over-Zealous Researchers May Be Setting the Stage For Yet Another Colossal Failure
Towards the beginning of a recent conference on Surrogate Markers for HIV Infection held in Arlington, VA, Donna J. Freeman, Deputy Director, Division of Antiviral Drug Products, at the Food and Drug Administration invoked Lewis Carroll's Alice in Wonderland:
Alice: Which way should I go? Cat: That depends on where you are going. Alice: I don't know where I'm going. Cat: Then it doesn't matter which way you go!
Dr. Freeman's quotation from the Victorian novel was meant as a warning for those assembled to clearly define the terms of the discussion on surrogate markers. Little did Dr. Freeman know that both scientists, activists and regulators had already passed "through the looking-glass," and that current debate on surrogate markers in HIV disease would resemble the Mad Hatter's Tea Party.
In testing therapies for HIV infection, one wants to know if a new treatment or therapeutic regimen can slow progression of HIV disease or prolong survival compared to no treatment or to the current standard-of-care. One of the major topics under discussion in the AIDS clinical research community right now is the use of newer tests of levels of HIV levels in the blood, particularly HIV-RNA (for "ribonucleic acid") in evaluating the efficacy of treatments for HIV.
From early on in the epidemic, many scientists and activists hoped to find a "surrogate marker" to substitute for disease progression and death as measures of the effects of anti-HIV therapies. In other diseases, such as cancer and heart disease, researchers have used tumor shrinkage and reductions in cholesterol levels, respectively, to serve as "surrogate markers" for assessing the clinical benefit of new drugs. Surrogate markers offer a way to expedite clinical trials of new agents by allowing researchers to observe and measure events that occur before, but predict later disease progression or death. Instead of seeing if people live longer by taking a new anti-neoplastic agent or heart disease medication, the investigator simply measures whether the new drugs can shrink tumors or lower cholesterol.
In AIDS, investigators traditionally look to see if antiretroviral agents raise the numbers of CD4+ T-cells in the blood. In each of these three diseases, the choice of "surrogate markers" seemed to make common sense. If in a person with cancer, a drug shrinks the tumor; if in a patient with coronary heart disease, a drug lowers the cholesterol level; if in a patient with HIV, a drug raises the level of circulating T-cells, one could logically conclude that these drugs would help the patient to live a longer and healthier life. Right? Wrong. In each of these disease scenarios, including HIV infection, beneficial changes in "surrogate markers" have not been able to reliably predict the clinical benefit of new therapies. In other words, in some cancer trials tumors have shrunk, in some heart disease trials cholesterol levels have dropped and in some AIDS trials CD4+ T-cell numbers have risen, but the therapies that have caused these "surrogate marker" changes haven't extended disease-free time or survival. How is this
possible? The pathogenesis of many diseases is complicated, and isolated features of the illness under consideration may not tell the whole story of what is happening in the body. Hence, "beneficial" effects of drugs on one or even several features of a disease may not lead to longer periods without symptoms, or a longer life.
"Viral load" tests are the new candidates in the AIDS clinical research community for use as surrogate markers to evaluate drugs for HIV infection. Many researchers (and some AIDS activists) want to use viral load assays to decide whether or not new therapies for HIV infection can extend AIDS-free time or survival. After all, it makes common sense that in a viral disease such as HIV infection if we can reduce levels of virus in the blood, people will live longer and healthier lives.
Common non-sense. The only way to assess if a drug prevents sickness and death is to show that it prevents sickness and death. Several years ago, many researchers (and most AIDS activists) insisted that small, short-lived, drug-induced changes in CD4+ T-cell numbers would predict the clinical benefits of antiviral therapy. After all, looking at changes in CD4+ T-cell numbers was a "direct measure of immune damage." In fact, in a recent review of 16 trials of antiretroviral therapy in AIDS by Dr. Thomas Fleming of the University of Washington, positive changes in CD4+ T-cell numbers correlated with clinical benefit in only about half the cases. Small, short-lived, drug-induced changes in CD4+ T-cell numbers are about as predictive of clinical benefit as tossing a coin.
Viral load changes may indeed be predictive of the clinical effects of antiretroviral therapy. In the current discussions, there are those who insist that viral load assays are a "direct measure of viral activity" and therefore changes in viral load, by common sense, will have to correlate with the clinical benefit of drugs for HIV infection. Then, there are others who are not yet ready to make that leap of faith and are asking for large-scale studies to assess the ability of changes in viral load to predict the wellness and survival of people with HIV. There is a danger that those with a priori notions about the usefulness of viral markers may be tempted to "torture the data until it tells them what they want to hear."
Even now, although retrospective evidence has not reliably shown a correlation between viral load changes and clinical benefits of therapy, there are those who are waving around dubious analyses of single studies asking that viral load assays be incorporated into clinical practice and as primary endpoints in trials of antiretroviral therapy.
Why are some researchers and activists rushing to crown HIV RNA as the new surrogate marker of choice with little evidence to support their claims? First, many confuse a "surrogate marker" of drug effect with a marker of disease progression. For example, everyone would agree that a progressive and substantial decline in CD4+ T-cell numbers in predictive of actual disease progression and risk of death. However, small, short-lived changes in CD4+ cell numbers do not reliably predict the clinical benefit of drugs for HIV infection.
In other words, if your CD4+ cells drop from 500 to 5, you can assume you are getting sicker, but if you go from 499 to 524 T-cells for six months after taking AZT, you cannot necessarily assume you are going to live any longer than if you had never taken any antiretroviral therapy at all. While some studies are showing a correlation between disease progression and viral load (the sicker you get, the more virus seems to be around), many are jumping to the conclusion that dropping viral load a log or two, and for a short time, with a new treatment or treatment strategy is going to translate into clinical benefit.
Second, frustrated scientists have a new toy and biotechnology companies have a new assay to sell. A senior official from the National Institutes of Health recently remarked that he thinks that "the new technologies are driving this debate." In other words, since we have new and more sensitive tools to measure levels of virus in the blood, we are investing a lot of time and money putting our new capabilities to that very use. The companies that have developed these assays also have heavy financial incentives to get them into use and on the market as soon as possible.
All of this could have grave implications for people with HIV. Rushing to use surrogate markers to evaluate therapies for HIV infection pushes us away from getting the fundamental answers about the therapies people with HIV are putting into their bodies. After some ten years of antiretroviral research, we can confidently say that antiretroviral therapy raises T-cells, but we have scant evidence if any of the four drugs currently approved used alone or in combination (except for AZT in advanced disease) can prolong survival!
Dr. Donald Abrams from San Francisco General Hospital and University of California has said in public meetings on numerous occasions, "I'd rather have my friends alive than to have them have 20 more T-cells." In a few more years, is Dr. Abrams going to have to revise his comments, after we have a rigorous profile of the ability of the protease inhibitors to depress viral load but little clinical information on their effects, and say, "I'd rather have my friends alive than to have them have a log less virus in their blood?" The truly agonizing feature of this debate is that there is a solution sitting right in front of our noses, but vested interests and small and dogmatic minds in both the scientific and activist communities have conspired to push us to make the same mistake we made in believing changes in CD4+ T-cell numbers predicted clinical benefit instead of proving it. Large-scale studies of the protease inhibitors, the latest class of antiretrovirals, using perhaps as few as 5,000 patients, could tell us if these
new drugs prevent sickness and death and if the new viral load assays are useful predictors of the clinical benefits of these drugs.
We need larger studies because the less powerful a drug is, the more people you need to study in order to show its effect. With the nucleoside analogues, we over-estimated their potency, designing trials for homerun drugs, leaving us with studies lacking the statistical power to tell us if we even had a base hit. The Pollyannas among us are running around saying the protease inhibitors are homeruns and that we don't need big trials. The drug companies want to do drug development on the cheap and to sponsor small studies which are unlikely to tell us about the effects of these drugs, unless they are indeed "grand slam" agents.
Scientists and activists addicted to smaller studies still believe that these "statistically weak" studies are going to be able to reliably give us the answers about the clinical benefits of mediocre drugs and to validate the new measures of viral load. Maybe one day we'll do things the right way in AIDS research. With the new viral load markers, though, it appears we're less inclined to learn from recent history than we are to repeat it.