Better Viral Tests: Interview with Mark B. Feinberg, M.D., Ph.D.

Background

Accurate blood tests for HIV activity, and for other markers
of disease severity or progression, are critically important
for developing new drugs. Reliable tests might shorten the
time required to show which drugs are good candidates from
years to months, allowing many more potential treatments to
be tested. And better viral tests could also improve medical
care with the drugs we already have, by showing when a course
of treatment is working for an individual and when it is not,
so that the physician will have rational guidance on when to
add or to switch therapies.

But what is happening today in the development of viral
markers is poorly understood in the AIDS community. People do
know that the tests we have now -- T-helper count as a
measure of immune function, and p24 antigen as a measure of
viral activity -- have serious drawbacks when used to measure
antiviral drug effects. But many do not know how much
progress is being made in developing better tests, both for
drug trials and for clinical care. These tests are not yet
commercially available to physicians, but some of them may
become available in the not-distant future.

Scientists, physicians, and activists have sometimes drawn
the wrong conclusions from the atmosphere of gloom which
followed the Berlin conference -- suggesting, for example,
that we cannot rely on laboratory tests in trials of
antiviral drugs, and therefore must use only "clinical
endpoints" -- death or major disease progression -- for
trials of antivirals. Such a policy would add years to the
time required to test potential drugs. And when a drug fails
to work, clinical endpoints do not tell why.

In fact, researchers are now learning much more than was
known before about HIV pathogenesis, about how to measure it,
about why past trials have often failed to provide the
information physicians need, and about how to run better
trials in the future. To focus on the new developments in
antiviral testing, which are part of this expanding
knowledge, we interviewed Mark B. Feinberg, M.D., Ph.D., a
scientist at the Gladstone Institute of Virology and
Immunology, Director of the Virology Research Laboratory at
San Francisco General Hospital, and Assistant Professor of
Medicine, Microbiology and Immunology at the University of
California, San Francisco. (The Gladstone Institute, a major
new research center, is associated with San Francisco General
Hospital but largely funded by a private charitable trust.)

Glossary: Viral Tests

This section explains some of the viral tests discussed
below, which will be unfamiliar to many of our readers.

* PCR (polymerase chain reaction). This is a test which
detects very small amounts of a specific kind of DNA -- the
molecule which contains the genetic code of most living
things. In nature, DNA is able to reproduce itself inside
living cells; this is necessary for life to be able to
reproduce. PCR creates conditions in the test tube so that a
particular sequence of DNA (if it is present in the sample
being tested) can be reproduced; each strand will become two
strands. The process is repeated, and the two strands become
four, then eight, then 16, etc. After 20 or more cycles of
this process, each molecule of the sought-after sequence of
DNA will have produced a million or more copies -- enough to
be easily detected by standard methods, even when the
starting concentration was far too low to be detected
directly. (The inventor of PCR, Kary B. Mullis, Ph.D.,
received a Nobel prize for the discovery last month.)

[Note: since HIV is a retrovirus, it does not contain DNA;
instead its genetic information is in the form of RNA. A PCR
test for HIV usually looks for the RNA of the virus. Before
the PCR can be run, the RNA is transcribed into DNA, using an
enzyme (reverse transcriptase) which HIV itself uses for this
purpose.]

* Quantitative PCR. Originally, PCR could only tell if a
particular sequence of DNA was present in a sample or not --
it could not tell how much was there. It is difficult to
accurately tell how much, because the successive steps of DNA
reproduction are not 100 percent efficient, and errors
accumulate throughout the process.

Today, a number of methods of quantitative PCR are being
developed. But it is not clear how accurate some of the
methods are. Errors of several hundred percent have been
common.

* QC-PCR (quantitative competitive PCR). This is a method of
quantitative PCR which was developed by Genelabs Inc. of
Redwood City, California, and published in Science on March
19, 1993. In QC-PCR, a known, tiny amount of a control
sequence, similar to the DNA sequence being tested but able
to be distinguished, is added to the sample before the PCR
process begins. During the successive doublings, both the
target sequence (the one being looked for) and the control
sequence, are multiplied similarly. Finally, the quantity of
the target sequence in the original sample is calculated from
its ratio with the control sequence. This is more accurate
than measuring the total amount of the target sequence, since
the inevitable errors in the doubling process affect both the
target sequence and the control sequence equally.

Unfortunately, QC-PCR is labor intensive and difficult to do.
Therefore it will not be available for routine medical use in
the foreseeable future. But it can be used in clinical
trials, and can also serve as a "gold standard" by which
other tests can be measured.

* Branched DNA assay (bDNA). This test, being developed by
Chiron Corporation, is much easier to run than QC-PCR and
gives similar results, except that it will not work at all
when the level of virus is very low.

In this test, copies of a DNA probe are attached to the wall
of a small laboratory vessel; then the sample is put in. The
probe binds to a certain part of HIV RNA, if it is found in
the sample, holding the RNA in the vessel. Then another DNA
probe is put in; one end of this attaches to another part of
the HIV RNA. The other end of this second probe has many
branches, and each branch ends with a "reporter" chemical
which, under certain conditions, will produce light, which
can be detected by laboratory equipment. Each molecule of HIV
RNA can attach to one of these branching structures and hold
onto a number of the light sources, not just one. In this
way, very small amounts of the target RNA can be detected,
without the need for PCR.

* P24 antigen test. This blood test has been used in medical
practice for several years. But recent data is showing that
it does not work well to predict clinical outcome. The reason
seems to be that the test is often inaccurate; people can be
p24 negative and still have large amounts of infectious virus
in their blood.

A new version of this test, called the acid-dissociated (or
ICD) p24 antigen test, offers some improvement. But it
appears to have most of the problems of the original p24
test.

* Viral cultures. In this test, virus is grown from blood
cells or blood plasma. To estimate how much virus is present,
the sample is successively diluted until no virus is found.

Quantitative viral-culture tests have been used in research
for several years. But they are labor intensive and require
special facilities, so they have seldom been used in clinical
care. Also, they may be less accurate than the newer tests.


The Interview

JJ: What is your main work at the Gladstone Institute?

MF: Our Virology Lab (at San Francisco General Hospital) is
responsible for doing virologic characterization of
volunteers in drug trials. But our main work at Gladstone is
guided by our own interests in better understanding the
natural history of HIV disease. We need to understand the
natural history to determine whether any intervention is
beneficial or not. If you don't know what correlates with
disease progression, you're in the dark.

Over the past couple of years, a better understanding of HIV
disease has emerged. There is more viral replication going
on, even during the asymptomatic phase, than people had
suspected. Also, clinical deterioration is presaged in many
ways by evidence of increased viral replication -- as if the
immune system of the infected person is losing its ability to
contain viral growth. What may be going on is that as the
virus replicates more, it can inflict more damage on the
immune system, which consequently further weakens its ability
to hold the virus in check.

We, and others, are trying to find ways of accurately
monitoring these processes. We certainly need a better
baseline description of what's going on in people with HIV.
We need to nail down the association between increased viral
replication and disease progression. We also need accurate,
practical markers for drug trials. The tests that have been
available, and the ones that are commonly followed in drug
trials, are not very good.

The ideal would be a test that is accurate, that correlates
with the fundamental biology of the disease, that predicts
future disease progression and the effects of treatments, and
that is doable on a large scale. Then clinicians in practice
could monitor their patients, and get insight directly into
what is going on, instead of relying on surrogate markers,
which have obviously been incompletely helpful.

JJ: You would not describe a good viral test as a surrogate
marker?

MF: Right, it's a direct marker. Surrogate markers are by
definition things other than what you want to measure.
Surrogate markers give you a clue about what's going on. But
they are only meaningful if they accurately reflect the
underlying process. Certainly CD4 counts (T-helper counts) do
to a certain extent, but what controls the level of CD4 cell
is not entirely understood. Part of it could depend on virus
replication, part of it could be disruption of the
architecture of the immune system. There are many variables,
many unknowns, which may affect CD4 counts. People used the
CD4 test because it was available, and it seemed like it
would be a meaningful marker. Early on it showed some
response to therapy. In retrospect, in more detailed
analysis, it has not turned out to be that good.

Clearly CD4 is useful for staging people, for knowing when to
begin prophylaxis for opportunistic infections. Here it is a
meaningful marker. And falling CD4 counts clearly correlate
with development of immunosuppressive disease.

JJ: Concerning the p24 antigen test, are its problems caused
by the fact that much of the virus is complexed with
antibody, so it isn't detected -- even with the newer acid
dissociated (ICD) p24 antigen test?

MF: Yes. We have long had the sense that was true. But you
get a much clearer picture when you have other ways of
looking for the virus that's there.

There are a number of proteins, or other parts of the HIV
virus, that one could conceivably measure, if we had tests to
do so. P24 was the first test that was developed. For many
purposes it's useful. For example, as a laboratory measure
for monitoring virus replication in tissue culture, it is
very good, because there are no antibodies to HIV in those
tissue culture flasks. The test is very sensitive, very
straightforward to do, and it is accurate in that context.
But in the body, antibodies complicate the detection of p24.

Another kind of entity you can look for, when testing for
HIV, is the RNA genome. HIV, being a retrovirus, has in the
virus particle two copies of its RNA, which can transmit the
genetic information of the virus to another cell. There are a
number of ways now -- two primary ways -- to look at the
amount of HIV RNA in the plasma of an infected person. One
uses PCR, the other uses branched DNA (bDNA) technology.

When you do those tests and compare their results with those
of the p24 antigen tests, you can find people who have very
high levels of HIV RNA in their plasma and no detectable p24.
So you have to conclude the p24 is falsely negative. In some
of those people, you can culture infectious virus from their
plasma, when their p24 antigen test is negative.

Part of the problem is that a p24 antigen result is not only
a function of how much virus is there. It is also a function
of how much antibody is there, and both the viral and
antibody levels could be changing, either together or
independently, in ways that we don't understand. This will
not necessarily be the same at each stage of the disease, or
for each individual. Probably the majority of people with
AIDS will not have detectable p24.

With the new methods, you can measure HIV RNA in the vast
majority of people.

Differences Between Tests

MF: We have tried to compare a number of different viral
markers head-on -- plasma culture, p24 antigen, QC-PCR, and
bDNA -- to see how they relate to each other, and try to
figure out which is the most meaningful one to follow. To
answer those questions, you need to do all the tests and
compare them directly, studying patients at different stages
in the infection, and doing all these assays on the same
specimens. This tells you how good the assays are, but it
also can tell you important things about the pathogenesis of
the disease.

JJ: Are you doing these tests as part of the trials run by
the ACTG (AIDS Clinical Trials Group, of the U.S. National
Institute of Allergy and Infectious Diseases)?

MF: We are doing this work independently, because we think
it's important. There is a frustration with the fact that
clinical researchers have not arrived at a consensus about
what they think is the best quantitative parameter to follow.
People are searching for better tools.

JJ: Where do the samples come from?

MF: Mostly they come from people participating in various
trials where extra blood samples are available, or from
people being cared for in the clinics here.

JJ: Is there a way somebody could sign up if they are
interested in donating blood and learning their results?

MF: Unfortunately we don't have a mechanism for that now.
Hopefully there will be more opportunities in the future, as
we organize trials that are based on viral quantitation as a
guide to therapy. We're working hard to make this available.

If we treat people based on their direct virologic measures,
could we do better than we've done by treating them based on
surrogate markers? My feeling is that the answer will be yes.
But it's essential that we answer that question as rigorously
and correctly as we can. If we falsely assume that something
is logical and therefore would be a good marker, and it turns
out not to be, a few years from now we're basically in the
same situation we are now with CD4 counts. My belief is that
measuring the viral parameters directly will be more useful,
more accurate, and a better way to monitor the activity of
drugs.

JJ: Beyond being sure that the new tests measure the virus
accurately, people also want to see clinical confirmation. Do
the tests help predict which patients will do well in the
future, and which patients are benefiting by their drug
therapy? Of course we don't want to wait many years to get
that information from people who are being tested now. Can we
speed this process of validating the new viral markers by
testing samples frozen years ago, so that the future clinical
course of the patient is already known?

MF: An advantage of both QC-PCR and bDNA is that they can be
done on frozen specimens, provided they were frozen in an
appropriate way. We may be able to learn a lot from trials
that have already been done.

Other research groups are finding that measuring viral burden
quantitatively is in fact predictive of disease progression.
It's not yet proven that intervening to decrease viral
burden, if that is possible, would slow down progression.
That's exactly the kind of information one needs. For a
number of ACTG trials, there are lots of stored specimens
that might be used.

There is already a fair amount of evidence that increasing
viral replication is an indicator that disease progression is
accelerating. What we don't yet know is whether, if you can
slow down the virus, you can slow down progression. It sounds
logical, but it does need to be demonstrated.

We do not know whether AZT alone will give a convincing
demonstration. And we do not always have the best samples --
for example, where you know how long somebody has been on the
drug, and whether they had resistant viruses.

JJ: After the Concorde results last summer (which raised
serious doubts about using T-helper counts as a measure of
antiviral efficacy in trials), some people have suggested
that we need to return to clinical endpoints for trials,
instead of using blood tests to determine how well a drug is
working.

MF: My feeling about the reaction to the Concorde results is
that people were understandably disappointed, but they
sometimes took the wrong lesson from it. I don't think it's
correct to conclude that it does not make sense to intervene
earlier in the course of the disease. Everything I know as a
basic scientist studying HIV pathogenesis tells me that the
time to intervene is earlier, if you had effective drugs.

What I take from the Concorde results is that, if a drug or a
test did not work, we need to know why. The results do not
condemn early intervention; they may condemn the drug or how
it was used. But nothing about that trial has swayed my
belief that intervening early would be beneficial if we had
good drugs.

The Concorde study left some people with the notion that we
don't have a good marker for antiviral drug activity, and we
need to follow clinical endpoints. That basically puts us
back to where were seven years ago. We can do better today,
and we must do better. Following clinical endpoints will be
essential to correlate with other markers, like direct
virologic markers. But clinical outcomes take time.

And a clinical endpoint does not tell you why a drug did or
did not work. It's 1993, and we have to learn to understand
these things. Then it will be important to have a consensus
on what's best to look for in trials.

My feeling is that this movement toward saying, "We need
clinical endpoints," is like people throwing their hands up
in frustration, some in despair and some in confusion about
what do to. Many of us who work in this area do have a fairly
clear idea about what might be done.

JJ: In which patients are QC-PCR and branched DNA
appropriate? My understanding is that the branched DNA test
is much easier to do, and feasible for widespread use, but it
won't work if the viral level is very low, because of
background noise in the test.

MF: That's right. But it will pick up virus in most people,
if not almost everybody, with a CD4 count of under 500. For
people over 500, it may pick up 70 or 80 percent. And for
those who cannot be measured, it may be good enough to know
that they're below the cutoff of the test, because they may
have relatively little risk of rapid disease progression
(because the viral level is so low).

JJ: Whereas with the p24 antigen test, if you're below the
cutoff, it doesn't mean very much?

MF: Right, you could have AIDS and have p24 below the cutoff,
and still have a hundred thousand, or a few million, copies
of the virus per milliliter of blood at that time.

The QC-PCR, or other quantitative PCR methods, are very
important; when we have learned enough to be sure of their
accuracy, these tests might serve as a gold standard for the
branched DNA or other assays. But quantitative PCR assays are
labor intensive, and demand high levels of skill from the
people doing them. It's hard to imagine that these will
become available enough that physicians would use them in
routine care.

The branched DNA assay could be widely used because it is
more like other clinical tests that are commonly done. Also,
it has a much more rapid turnaround time than quantitative
PCR.

But it is still too early to give a blanket endorsement to
the bDNA assay. It still needs rigorous testing, with
additional comparisons with other tests; that is what we and
others groups are doing. Hopefully we will have answers to
this in the relatively near future. But also, the test has to
be validated in clinical practice; that means doing clinical
trials using this test as a marker.

JJ: What is the range of the number of virus copies which you
find in patients' blood with these tests?

MF: In people who have just recently been exposed to HIV and
are undergoing the primary infection -- where the virus is
widely disseminated in their body, and there is no immune
response to it which has developed yet -- they can have
millions of copies of RNA per milliliter of blood. You divide
that by two to get the number of virus particles, since there
are two copies of RNA in each virus particle. So people can
have millions of virus copies per milliliter.

Once they have an immune response to the virus, the levels of
viremia go down significantly. But still people will have
10,000, 20,000, or 100,000 copies of the virus per
milliliter, and some people have more.

JJ: What are the limits of the tests -- the number of copies
below or above which they do not work?

MF: The cutoff for the bDNA assay is presently 10,000 copies
of RNA per milliliter. That is the background level that was
established after screening many HIV-negative plasma samples.
Researchers are working to lower that background.

The QC-PCR assay can go down to about 10 copies per
millimeter. And smaller amounts can be detected by special
preparation of the sample. If it is done well, there's no
background. So QC-PCR is the best assay to determine whether
a sample is negative.

JJ: How high can these tests go?

MF: You can cover the whole range, to several million copies
per milliliter of blood.

JJ: Do you have any thoughts about how to get the new tests
quickly accepted? Presumably the first step is to finish
getting the data, and make a good scientific case.

MF: That's an essential first step. It is important that this
work comes from a number of laboratories. Because what will
matter is that you get the same result from the test no
matter where you do it.

After we make sure that the assays are accurate, which I
think we're pretty close to, we need to validate them in
clinical trials, to show that they make a difference (in
clinical outcomes). It would not help to spend money for
these tests if it turns out that monitoring them does not add
anything to clinical care of a patient. We must make sure
that monitoring these tests allows us to treat patients
better with the drugs that we have -- and also allows us to
identify effective drugs more expeditiously, and to eliminate
those that don't work.

JJ: How might these tests be used in clinical trials?

MF: Perhaps a valuable initial screening, which you could do
fairly rapidly, without necessarily needing a lot of people,
is to test a drug in HIV-infected people and see how well it
decreased viral replication. If it did, then in a short
period of time the drug could go on to larger studies, to
look at endpoints.

Many people have participated in large trials of AZT and
related drugs. These trials are expensive, and very demanding
on lots of patients who participate. And at the end of the
day we don't know clear answers from the studies. We need to
be more selective about what we do, to develop a more
productive framework to test drugs.

Later, if we prove in one trial, or in a few trials, that
diminishing the viral burden correlates with a better
outcome, there may be a day in the not too distant future
when viral reduction might be sufficient proof of efficacy
for the FDA. They don't yet know what criteria to settle on.
I hope that the data about direct virologic markers will be
persuasive enough that it will be easy for the FDA to decide
whether or not to use them.

There are a number of research groups now working on viral
tests to measure HIV disease progression; there is much work
in this area. The lessons I stated above were derived not
only from this laboratory, but from many other groups as
well.

For me, the important point is that we are not in the dark.
It's not like there is no course we can take that will give
us definitive answers. Unfortunately, that is the lesson that
many people derived from recent disappointments in
therapeutic interventions.

What everybody has to do, from physicians who take care of
patients, to the patients themselves, to the scientists, is
to tackle the difficult questions and devote energy to that.
There is important work that can be done today.

This should be a source of optimism. We will learn why the
drugs we have today don't work as well as we hoped. We may
also be able to learn how to use them better -- which would
improve patient care in the short term. And we will have a
better way of testing new drugs rapidly.