Testing for HIV mutations
hard to resist
Doing more than testing and switching
William Check, PhD
"New! Improved! HIV-1 Resistance Tests—Get One Today!"
Clinicians and laboratory scientists familiar with tests to detect
resistance to drugs used to treat human immunodeficiency virus type
1 would unanimously reject this claim as extravagant and overwrought.
Take away the "New!" and the exclamation marks, however, and they
would probably give it their stamp of approval, including the implication
that such testing should be routine for HIV-1-infected patients.
Greater acceptance and use of HIV-1 resistance tests rests not
on any single dramatic breakthrough but on many incremental improvements:
- accumulating data linking information from resistance tests
to clinical outcomes.
- introduction of several new antiretroviral drugs.
- approval of the first kit for genotypic analysis to detect resistance
- better algorithms to interpret sequence data from genotypic
- more biologically relevant cutoffs to interpret results from
But perhaps the strongest impetus to broader acceptance of HIV-1
resistance testing is an urgent need arising from growing rates
of resistance in HIV-1 infections.
"There continues to be an accumulation of studies showing an increasing
prevalence of drug resistance among newly infected individuals in
North America," says Daniel Kuritzkes, MD, associate professor of
medicine and microbiology in the division of infectious diseases
at the University of Colorado Health Sciences Center and co-director
of the Colorado AIDS Clinical Trials Unit. In one study, by Susan
Little, MD, of the University of California at San Diego, resistance
rates among newly HIV-1-infected persons were found to have quadrupled
from 1995-1998 to 1999- 2000, with 12 to 14 percent of isolates
in the latter period being resistant to at least one antiretroviral
drug and several percent showing resistance to more than one drug.
Dr. Kuritzkes calls these data "disturbing" because of the high
numbers and, more important, because they show that "patients in
treatment are not being effectively counseled about how to prevent
Sounding the same theme, David Hillyard, MD, director of molecular
infectious disease testing at ARUP Laboratories in Salt Lake City,
says, "HIV resistance testing is still in its infancy, as are clinical
correlations to validate this testing. Yet," he says, "there is
a growing realization that there is a real need [for these tests]
due to emergence of resistant strains."
Dr. Hillyard cites reports that from 10 to 25 percent of treatment-naive
persons have one or more mutations that may make them resistant
to an antiretroviral drug. "So there is a tremendous need for tools
to help guide physicians in this very difficult process of how you
initially treat a patient," he says, "and in the even more difficult
situation of a patient who has failed therapy and for whom you need
to make a choice among a limited set of remaining drugs."
The data validate this approach, says Susan Eshleman, MD, PhD,
associate professor of pathology and director of the HIV genotyping
laboratory at Johns Hopkins Medical Institutions. "A number of retrospective
studies in the literature support the use of resistance testing,"
she says. "And prospective studies are beginning to appear." Based
on those clinical trials, three large networks—two in the
U.S. and one in Europe—have recommended using resistance testing
in various clinical settings. "Resistance testing is becoming part
of routine management," Dr. Eshleman says.
Mark Wainberg, PhD, professor of medicine at McGill University
and director of the McGill University AIDS Center, agrees: "There
is a consensus that good clinical management of HIV-1-infected patients
requires access to resistance tests."
Use of resistance testing may expand further now that one genotyping
kit—Trugene from Visible Genetics—has received 510(k)
market clearance from the FDA and a second genotyping kit—ViroSeq
from Applied Biosystems—is under review. Dr. Kuritzkes calls
approval of Trugene "a major advance" for the field. "It has helped
to establish standards and to lend legitimacy to HIV resistance
testing as part of patient management," he says.
Having an approved test should make reimbursement more widely available. "As
tests become FDA cleared or approved, reimbursement becomes easier," notes James
Versalovic, MD, PhD, director of microbiology at Texas Children’s Hospital and
assistant professor of pathology at Baylor College of Medicine. "There is more
widespread acceptance among insurers, including the federal government. Multiple
CPT codes have been established."
In their current state, however, resistance tests provide
limited information. "Existing resistance tests are better at telling
you what drugs a patient won’t respond to than telling you what
drug you could switch to that would be effective," Dr. Hillyard
says. In the words of Graeme Moyle, MD, PhD, associate director
of HIV research at Chelsea and Westminster Hospital, London, "Resistance
test results today are probably more predictive of failure than
Part of the problem is a limited number of therapeutic options.
"Drug development in HIV is still a growth industry," Dr. Wainberg
says. "We are treating people for life, and there is resistance
out there to every available drug."
Another problem is the increased use of three- and four-drug treatment
regimens. "Progress in resistance testing is slower because, with
today’s complex regimens, it is quite difficult to sort out the
contribution of each drug to resistance," says Charles Boucher,
MD, PhD, clinical virologist at University Hospital in Utrecht,
Interpreting resistance mutations is "extremely complex and somewhat
unpredictable," Dr. Eshleman confirms. Mutations may work synergistically
or additively. Some mutations cause cross-resistance to other drugs,
while others may reverse resistance to other drugs or even cause
hypersusceptibility. As a result, she says, "[Genotypic] resistance
testing requires a commitment from the laboratory not only to the
technical aspects of the work but also to interpretation of results."
Quality commercial systems are available to help with this work.
"Whichever interpretive system you use should be current and needs
to be reviewed on an ongoing basis by a panel of experts," Dr. Eshleman
advises. "Every few months we learn more."
To improve technical quality, the CAP has instituted proficiency
testing for genotypic resistance methods. The number of laboratories
participating grew from 15 to 20 in 2000 to 56 in 2002. "We’ve really
jumped," says Dr. Versalovic, who directs the CAP Survey. "Published
studies have demonstrated an advantage to treating patients based
on [genotypic] data, so physicians are demanding it and laboratories
are implementing it."
While challenges to interpreting genotypic test results are obvious,
"It is not that easy to interpret phenotypic resistance assays either,"
says Lisa Demeter, MD, associate professor of medicine at the University
of Rochester School of Medicine. "There is a big controversy in
the field now about how best to define phenotypic resistance thresholds."
Taken together, the rise of complex resistance testing and the
increasing number of antiretroviral drugs make treatment by specialists
more common, according to Robert Schooley, MD, head of the infectious
diseases division at the University of Colorado Health Sciences
Center. "There is less treatment now of HIV-infected persons by
doctors who don’t do it most of the time," Dr. Schooley says.
Indications for resistance testing are well defined. Recommendations
concur that resistance data be obtained after failure of first or
subsequent drug regimens. "The patient who benefits most from resistance
testing would be one who has failed a previous drug regimen or regimens
but who still has treatment options," says Dr. Demeter. "For a patient
who has been treated with all current antiretroviral drugs and has
virus highly resistant to all those drugs, resistance testing is
unlikely to help."
Resistance testing is also appropriate for treatment-naive persons
who are diagnosed when acutely infected, because of the rise in
the transmission of resistant virus and the clinicians’ tendency
to treat individuals identified with early symptomatic infection.
"I think most experts would recommend that you test those individuals
but not delay treatment while you wait for results," she says. Treatment
can be modified if resistance data make it necessary.
"The only group in whom resistance testing appears not to be indicated
are chronically HIV-infected persons who have never been treated
with an ARV [antiretroviral drug]," Dr. Demeter continues, "and
there is even some controversy about this." Persons who are now
chronically infected were initially infected when rates of transmitting
resistant virus were much lower.
Dr. Moyle makes the opposite case, but cautiously. "A person may
rebound on their first-line therapy because they started with a
resistant virus strain," he says. "So there is a constructible argument
to test before starting a patient on therapy. Although information
is insufficient to say whether that is helpful."
Encouraged by official recommendations, use of resistance testing
has risen, but not dramatically. "There has been a steady increase
in the last few years in our clients ordering resistance testing,"
says ARUP’s Dr. Hillyard. "Still, a clear distinction should be
made between resistance testing and viral load testing."
While all physicians who manage HIV-infected patients order viral
load tests regularly, Dr. Hillyard observes considerable variation
in the ordering of resistance tests. Some clinicians, including
experts who are very astute in managing HIV disease, rely very little
on resistance testing, he finds. "It should be remembered that people
fail therapy for a variety of reasons," he says, "and only one part
of the problem is access to a high-quality HIV resistance test."
Dr. Hillyard also notes the opposite—some clinicians with
less expertise may rely too heavily on resistance testing. "They
are tempted to see resistance testing as a stand-alone means to
make decisions," he says, "while most clinicians would feel that
HIV resistance testing is only one tool to use in making treatment
decisions, and one that is in the early stages of proving itself."
In Dr. Boucher’s view, there has been "surprisingly slow pickup of genotypic
testing throughout Europe and the U.S., which is mainly I think a function of
reimbursement." He sees "a major disconnect" between published data and how
widely testing is employed. "One other reason resistance testing is taking off
slowly," he adds, "is the need to be an expert or full-time practitioner to
use it well."
Results of several randomized prospective clinical trials
that have been published or reported at meetings provide evidence
that genotypic resistance testing yields clinical benefit. Data
supporting a possible advantage to treatment guided by phenotypic
analysis are still sparse.
In genotypic analysis, resistance is evaluated by determining
the nucleic acid sequence of critical segments of viral genes for
the two antiretroviral drug targets—reverse transcriptase
and protease—and identifying mutations associated with resistance.
In phenotypic analysis, segments of reverse transcriptase and protease
genes are spliced into a standard HIV-1 genome and the recombinant
clone is incubated in vitro with antiretroviral drugs, similar to
assaying resistance to antibiotics. A medium-size to large molecular
laboratory can set up genotypic analysis or samples can be sent
to a reference laboratory. Specimens for phenotypic analysis must
be sent to Virco’s laboratory in Mechelen, Belgium, to ViroLogic’s
laboratory in South San Francisco, or to Specialty Laboratories
in Santa Monica, Calif., which performs a version of phenotypic
analysis developed by French firm Viralliance.
In the first two controlled trials of resistance testing, GART
and VIRADAPT, after failure of a first antiretroviral combination
regimen, viral suppression was significantly more frequent among
patients whose next regimen was selected using results of genotypic
resistance analysis. GART used a panel of three experts who interpreted
genotypic information in conjunction with a patient’s history. Clinicians
in VIRADAPT used a simple algorithm to interpret genotypic results
Similar benefit from use of genotyping was seen in the HAVANA
trial, in which genotypic analysis was interpreted by a software
package called RetroGram written by Dr. Boucher. "In HAVANA, there
was also a strong suggestion that adding advice from experienced
HIV physicians to genotyping provided further benefit," Dr Kuritzkes
In the VIRA 3001 study, guiding regimen changes by results of
phenotypic analysis—Virco’s Antivirogram report—also
improved clinical outcomes.
Specialists evaluate the degree of clinical advantage from resistance
testing differently. "Knowledgeable physicians understand that it
can be a valuable tool but it is just one small piece of the puzzle,"
says Dr. Hillyard. "So far the results are promising but somewhat
modest in terms of clear-cut efficacy."
On the other hand, Dr. Demeter says, "The magnitude of effect
you see in GART and VIRADAPT almost certainly underestimates the
extent of benefit that you would get from genotypic resistance testing
in clinical settings." Not all trials provided expert advice, she
notes, and some provided advice that we now know was less than optimal.
"So I think it was quite remarkable that we saw as big a difference
as we did," she says.
Dr. Boucher calls beneficial results to date "relatively small,"
which he attributes to extensive cross-resistance within antiretroviral
drug classes and to the interpretive challenge. "It is still somewhat
of an art to properly interpret genotypes," he says. Better interpretation
will be a function of better training and better software provided
by vendors, he adds. Results of ongoing clinical trials relating
genotypes to clinical outcomes will advance the interpretation software.
Optimizing these rich datasets in the laboratory will require
in-house expertise, Dr. Hillyard says. "We have, as do most laboratories
specializing in this area, an experienced HIV physician responsible
for keeping up with the rapid changes in HIV sequence databases,"
he says. He calls such work "incredibly complex."
Even experienced HIV clinicians consult among themselves on how
best to interpret genotypic results in terms of possible drug regimens.
Dr. Moyle participates in such a group at his hospital, and Dr.
Demeter’s practice group holds conferences every few months to discuss
therapy in problem cases.
Beneficial interpretation of genotyping data assumes that laboratories
have accurately determined the sequence. To advance this goal, the
CAP’s HIV Surveys now include with each mailing one HIV genotyping
challenge. Dr. Versalovic explains that each genotyping proficiency
sample contains heat-inactivated HIV-1 that has been isolated from
"We have several HIV-1 strains with different sequences," he says,
"and we use a single donor for each sample. The challenge for the
laboratory is to identify the predominant mutations that may affect
susceptibility of virus to antiretroviral drugs and to recognize
In the CAPHIV/HV2 Survey, laboratories are required only to report
whether there are important mutations in either the reverse transcriptase
or protease genes but do not have to specify the mutations. "Reporting
requirements will continue to get more rigorous," Dr. Versalovic
With both the CAP and private companies providing standards and
proficiency testing, there is now adequate proficiency testing available,
Dr. Versalovic says.
In addition to trials in which resistance testing proved beneficial,
there have been several in which it did not improve viral suppression.
In NARVAL, conducted in France, no clinical benefit was seen with
either genotypic or phenotypic analysis. Lack of benefit with genotypic
analysis might have been due to the heavily pretreated status of
many patients: In a post-hoc analysis, genotyping was beneficial
in patients who had failed fewer previous regimens.
Even in this subgroup, however, phenotyping was not advantageous.
Failure of phenotyping in NARVAL and in the CCTG 575 study, which
employed ViroLogic’s Phenosense system, could have been due to inappropriate
cutoffs for some commonly used nucleoside reverse transcriptase
inhibitors, particularly dideoxyinosine (ddI), stavudine (d4T),
and abacavir. In these studies, a 2.5-fold increase in IC50
was used as the definition of resistance. But in a retrospective
analysis, treatment outcome was predicted only by a much lower cutoff,
1.6- to 1.7-fold. Such small shifts are difficult to measure, according
to Dr. Kuritzkes.
"You are getting very close to the technical reproducibility of the assay,"
he notes. "These cutoffs may yield significant effects in the overall study
population, but for an individual patient 1.7-fold could be 1.5-fold or 1.9-fold."
Estimates of the relative value of genotypic and phenotypic
analysis vary. Dr. Schooley sees genotypic testing as generally
more difficult to interpret. "You have to understand the predictive
value of specific mutations," he says. While most laboratory reports
state what genetic changes typically mean, he notes that these interpretations
rely on algorithms, either based on rules selected by expert panels
or generated by computerized decision trees. "Clinicians can work
through these algorithms," Dr. Schooley says, "though it may take
a few times to become conversant with them. And it can be helpful
to have someone available who is experienced."
Phenotypic testing, on the other hand, "is analogous to what we
have been doing for a long time with antibiotics," Dr. Schooley
says. "It takes less time to assimilate."
To other specialists, however, the distinction is less clear-cut.
Expressing a more common view, Dr. Demeter says genotyping is better
at picking up early mutations that may signal evolving resistance.
Also, she notes, "Phenotyping may be less sensitive at detecting
resistance to ddI and d4T." In heavily pretreated patients who may
have complex mutation patterns, phenotypic testing may offer an
"But in patients with limited experience, I really don’t see any
advantage [to phenotypic analysis]," Dr. Demeter says. "Phenotyping
is not the preferred initial test. If you look at prospective clinical
trials, they provide stronger support for use of genotypic testing.
So even if the cost of the two methods was similar, genotyping would
have a slight edge." In fact, phenotyping is more expensive than
Nor does Dr. Demeter think that interpreting results of phenotypic
analysis is more intuitive. "I don’t think a phenotype is that much
easier to interpret than a genotype," she says. "It is not completely
straightforward. We still don’t fully understand how best to define
Dr. Versalovic says the subset of patients in whom phenotypic
analysis may be advantageous is not large. "When I was at the Massachusetts
General Hospital," he says, "we reviewed the data and found that
phenotyping would have helped in 10 percent or fewer cases. Resistance
is a moving target," he acknowledges, "and with time there may be
a greater need for phenotyping as patients are on treatment longer
and more agents are used for treatment. But right now it is hard
to justify the cost as a routine test."
Dr. Versalovic also questions whether it is really so easy to
interpret phenotypic data. "Where do you set the cutoff?" he asks.
"The IC50 for each drug, although based on correlative
data, may be arbitrary." Cutoffs have been modified for some drugs
based on accumulating data, Dr. Versalovic says. "Clinicians often
prefer phenotypic analysis because the results appear easier to
interpret," he continues. "But what does that mean in the patient?"
Dr. Moyle agrees. "Phenotypic analysis doesn’t always give us
better numbers that we can more easily draw conclusions from." He
notes that cutoffs can be of three kinds: a technical cutoff, based
on assay reproducibility—is the tested virus significantly
different from the reference virus?; a biological cutoff—is
the patient’s isolate different from what we would find in a drug-naive
population?; and a clinical cutoff—is the tested virus likely
to respond to the drug we have tested against? Clinicians need to
know the clinical cutoff. However, Dr. Moyle notes, "There are only
a few drugs for which that is established." For other agents, the
biological cutoff is reported.
But just as there are few drugs for which clinical cutoffs have
been determined for phenotype, there are few data correlating the
presence of specific mutations with loss of antiviral activity,
Dr. Kuritzkes emphasizes. "In most cases," he says, "the significance
of resistance mutations is judged by their correlation with phenotypes."
Whatever the advantages and disadvantages of each method, clinicians
are voting with their order slips.
"When I survey treating physicians in large audiences at clinical
meetings," Dr. Eshleman says, "I get a large show of hands for physicians
who have ordered genotyping and a scant show of hands for those
who have ordered phenotyping."
The ordering at ARUP would bear that out. "In our laboratory we have by far
the greatest subscription to genotypic testing," Dr. Hillyard says. He attributes
this to genotyping being more affordable and more rapid. Clinical data don’t
overwhelmingly favor either method, in his view.
With approval of Visible Genetics’ genotyping kit and the
possible future approval of Applied Biosystems’ kit, competition
in this market should intensify. Is either product superior?
Dr. Demeter has experience in her research laboratory with both
ViroSeq and Trugene. Both are robust and produce clean sequences
from clinical specimens and both are labor-intensive. She sees a
throughput advantage to the AB system. "But Trugene’s throughput
is more than adequate for what most hospitals would require," she
"There is no clear winner in my opinion," says Dr. Versalovic,
who is now setting up a new molecular diagnostics laboratory. "Both
are quality systems for clinical use." One advantage of the Visible
Genetics product is that it is FDA cleared. On the other hand, he
says, "Technically I prefer AB’s capillary-based system."
Assuming that AB’s product is soon approved, Dr. Hillyard sees
a problem that FDA’s way of setting product standards will pose
for the interpretive software of both systems.
"Let’s say we have two FDA-approved tests," he hypothesizes. "Knowledge
about resistance mutations is changing rapidly. How will their algorithms
be updated?" He hopes the FDA will work out a process that won’t
require full clinical studies every time an interpretive algorithm
HIV resistance testing will also be at the forefront of an FDA
regulatory initiative regarding the large volume of molecular testing
in clinical laboratories performed without FDA-approved kits. "How
aggressive will FDA be in saying you must use approved tests?" Dr.
Virco offers a combined genotypic-phenotypic resistance analysis
called Virtual Phenotype. In this procedure, sequences from a patient
isolate are matched against a database containing 28,000 patient
isolates on which Virco has done genotypic and phenotypic analysis,
yielding a probabilistic likelihood that the patient’s virus is
susceptible to each drug.
"I think the Virtual Phenotype is a clever and potentially useful
way to evaluate genotyping," says Dr. Kuritzkes. "Some studies,
most recently at this year’s retrovirus meeting, found similar outcomes
in patients managed by Virtual Phenotype as compared to patients
managed by actual phenotyping." These studies were not large enough
to establish true equivalence, he notes, but were probably accurate
to a first approximation.
Virtual Phenotype’s advantage is that it is less expensive than
actual phenotyping. A drawback is that it is not possible to match
for every base pair in reverse transcriptase and protease, so an
artificial intelligence program decides which codons to include.
"Virtual Phenotype is most helpful when there are a large number
of matches in the database and the distribution [of susceptibility
or resistance] among those matches is clear-cut," Dr. Kuritzkes
says. If, on the other hand, the database contains only 300 matches
to a patient isolate and only 60 percent of the matching genotypes
are sensitive (or resistant) to a particular drug, "in that case
it’s more of a coin toss," Dr. Kuritzkes says. Virtual Phenotype
provides all that information, so laboratorians and clinicians can
look at it critically and decide whether to rely on it.
Virtual Phenotype is offered through Virco’s U.S. partners—LabCorp,
Quest, and ARUP—which perform genotyping and send the sequence
to Virco in Belgium for the VP analysis. An individual physician
or medical center can also submit a sequence directly to Virco for
VP analysis through a service called VircoNet.
Dr. Demeter’s clinic gets Virtual Phenotype as its standard resistance
method through LabCorp. "It is a good test, and clinicians like
it because it provides a genotype and an estimated IC50,"
she says. Whether they will continue to rely on VP now that Trugene
has been approved has not been decided.
Tim Alcorn, PhD, director of the infectious disease laboratory at the LabCorp
Center for Molecular Biology and Pathology, says ViroLogic also has a combined
genotyping-phenotyping product called Phenosense GT, although it is not the
same as Virtual Phenotype. "Using genotyping and phenotyping together probably
gets you the best quantitative picture," Dr. Alcorn says, "although there are
no data right now saying that having quantitative results allows better treatment.
Qualitative information about reduced susceptibility seems to be as good."
While resistance testing can optimize regimen switches,
it can’t create new possibilities when a patient’s isolate is resistant
to all available drugs. Hence the continual search for better antiretroviral
"The challenge is to find new drugs with favorable pharmocokinetics
or that are not susceptible to cross-resistance within drug classes,"
Dr. Eshleman says. She notes that resistance is not all-or-none
but has to do with inhibiting viral replication at a particular
plasma drug level. A new regimen may overcome resistance because
of enhanced activity related to pharmacokinetics and attainable
Says Dr. Schooley: "We now have a better idea about so-called
cutoffs very much like what we see with antibiotics in bacteria.
A result on an agent in vitro needs to be interpreted in terms of
the amount of drug that will be delivered to the site of infection."
One prominent example is that trough levels of protease inhibitors
achieved in blood when these drugs are given as single agents may
be only slightly above the susceptibility of wild-type viruses.
However, trough levels are enhanced significantly by adding a second
PI, most often ritonavir. "This allows you to use these agents in
situations in which phenotypic susceptibility testing suggests they
might not be useful," Dr. Schooley says.
Another advance has been the generation of correlations between
resistance test results and clinical outcomes in the development
programs of three recently approved antiretroviral drugs, the protease
inhibitor ritonavir and the reverse transcriptase inhibitors tenofovir
and abacavir. Clinical evaluation of these drugs included intensification
studies in which the drug was added to a failing regimen, providing
an estimate of the impact on viral suppression of phenotypic or
genotypic resistance test results obtained at the time the drug
was added. Based on such studies, Dr. Moyle notes, package inserts
for these three antiretroviral drugs present suggested cutoffs for
phenotypic test results and algorithms for establishing likely efficacy
based on genotypic analysis. "This information can guide clinicians
in making decisions about the likely utility of those drugs," he
This advance fits with the big push to define resistance based
on clinical response. Says Dr. Kuritzkes, "It will be expected now
for new drugs that come along, particularly for drugs being developed
for use in salvage therapy, to have these correlations." He sees
manufacturers of previously approved antiretroviral drugs now going
back to find datasets to establish such correlations.
Dr. Kuritzkes points to a new class of drugs called entry inhibitors
that are under investigation. Such agents "have great potential
for improving salvage therapy," he says. For one of these, a fusion
inhibitor called T20, phase III studies are nearly complete. Entry
inhibitors interact with the HIV envelope protein, which is encoded
by a large gene, with resistance mutations scattered across it.
For this reason, Dr. Kuritzkes speculates, "It may be that phenotyping
will have an edge over genotyping in determining susceptibility
to entry inhibitors." At the retrovirus meeting, ViroLogic showed
data using Phenosense to detect resistance to entry inhibitors.
Virco is also developing a phenotyping assay for these agents. For
some newer entry inhibitors, there may be differences in susceptibility
even among patients who have never been exposed to them, so it may
become important to use resistance testing before starting therapy
with these agents, Dr. Kuritzkes says.
Finally, Dr. Eshleman has her eye on a laboratory enhancement.
"One very interesting area in the future will be combining resistance
testing with pharmacokinetic testing," she says. This will entail
not only determining the patient’s viral phenotype or genotype but
actually getting drug levels in a patient to determine what drug
level suppresses replication in that person.
"In this way we will get a better idea of the in vivo relevance
of resistance data," she says. And that, of course, is the purpose
of the whole exercise.
William Check is a medical writer in Wilmette, Ill.