For HIV genotyping, plenty of
  growing pains

May 2005
Feature Story

Something odd often happens after clinicians order viral load tests for their HIV-infected patients at William Beaumont Hospital, Royal Oak, Mich.

Or rather, what’s odd is what doesn’t happen: Few order a genotype test to assess drug resistance.

Carol Holland, PhD, technical director of molecular pathology at the hospital, says her laboratory performs approximately 30 HIV genotypes a year. When she worked at nearby Henry Ford Hospital, doing 15 to 20 such tests weekly was the norm. Granted, Beaumont doesn’t have the same patient population as Henry Ford, which is in Detroit. Nevertheless, Dr. Holland’s laboratory is fairly big and runs, by her estimate, a couple hundred viral loads a month. The number of genotypes should not be far behind.

Genotyping has become standard of care for many HIV-infected patients. The push isn’t coming from a new set of guidelines. The earliest recommendations suggested resistance testing could be used in virtually any situation, and that still holds true. Not all categories merited equal billing, however. Early on, resistance testing was strongly recommended for patients whose antiretroviral therapies were failing; less importance was attached to resistance testing for treatment-naïve patients. What’s emerging now, say observers, is parity among the classes.

Physicians have begun to make genotyping routine for patients even before they begin therapy, for example, as evidence grows that drug-resistant viruses are being transmitted sexually or through IV needle abuse. Reports suggest that anywhere from 10 to 20 percent of newly infected patients are infected with drug-resistant virus, while earlier estimates put the number closer to 10 percent or less.

That may come as news to some. For others, the news isn’t that such transmissions are occurring, but rather that these resistant viruses are more fit than previously thought. They’re not being overgrown by wild-type viruses, as some had predicted. "Now we’ve changed that thinking," says Mark Wainberg, PhD, professor of medicine, McGill University, and director, McGill AIDS Center, Montreal. "Because in fact many of the mutations hang around a lot longer than had previously been assumed."

Marie Hoover, PhD, sees other reasons for clinicians’ interest in genotyping, including what she calls the let’s-not-use-up-all-of-our-bullets-at-one-time stance. "They’re realizing the drugs are limited, and that by switching drugs every time they see the virus start going up, they’ve run out of drugs for some patients. So they want to see the genotype before they make therapy changes," says Dr. Hoover, who is CEO and lab director, Advanced BioMedical Laboratories, Cinnaminson, NJ. On the flip side, she encounters trigger-happy clinicians who want to change the therapy the minute they see an increase in viral load. Genotypes can stop an unnecessary switch in these cases: If there’s no evidence of resistance, the therapeutic failure is probably tied to noncompliance.

But even as the calls for resistance testing grow louder, they’re not always heard. "There aren’t nearly enough genotypes being requested by treating clinicians," says Dr. Wainberg.

Dr. Holland notes that her pathology residents know when a genotype is in order—when she trains them, she makes sure they understand when it should be used, when it’s not appropriate, and how to interpret the results. "But except for our ID physicians, most of our medical staff really doesn’t know," she says.

The obvious question to ask is, Why this gap? But given the mazy ways of HIV resistance, what’s remarkable is that there aren’t more.

It’s not unusual for a new test or technology to have a steep learning curve. As the method itself evolves, it usually becomes easier. It’s also not unusual for users to catch on as they gain experience with the new method.

Not so with HIV genotyping. Knowledge has remained in the hands of relatively few. "Clinicians who live and breathe this are more comfortable doing the interpretations themselves," says David Hillyard, MD, associate professor of pathology, Department of Pathology, University of Utah School of Medicine, and director of molecular infectious disease testing at ARUP Laboratories, Salt Lake City. Those who treat HIV patients have to be familiar with an extremely complicated pharmacology, thanks to the wide number of drugs and various interactions. For this group, "I’m sure there is more clarity about resistance testing than there was even several years ago," says Dr. Hillyard. He also concedes that reporting of resistance has perhaps become slightly more clear in recent years. Nonetheless, the vast majority of clinicians do not have sophisticated knowledge of mutations, how they work, nor what their impact is.

For them, HIV genotyping may always be somewhat of a Sisyphean adventure.

HIV is a swarm of different, constantly evolving viruses, and resistance changes all the time. A genotype is a mere snapshot into one little window of time, showing the dominant virus seen and sequenced at that moment. There will always be unknown mutations that can’t be accounted for, and there will always be unpredictable interactions among mutations.

Physicians work from a current list of mutations that have been identified as being predictive of resistance. But who knows how long the "real" list is? "It’s still kind of primitive," says Larry Reimer, MD. "We’re looking at a short list right now, when in reality there conceivably is a long list." Or maybe there’s no list at all. Perhaps the collective structure of the virus total is what matters, and not these individual site mutations, suggests Dr. Reimer, director of the microbiology laboratory and head of the section of infectious disease at the VA Medical Center, Salt Lake City; medical consultant to the HIV resistance testing laboratory at ARUP; and associate dean for curriculum and graduate medical education, University of Utah School of Medicine.

Genotype reports vary somewhat, depending on which system the laboratory is using. There are two FDA-approved systems, the VGI/Bayer TruGene and the Abbott (ABI/Celera) ViroSeq, in addition to home brews. At this point, the various systems have much in common. Says Dr. Hoover: "Really and truly, everyone’s sequencing everything, the same regions of the HIV." But not everyone reports the same data. That’s where the algorithms come in.

The interpretation of a genotype report is driven by algorithms—a different one for each system. And even though "there’s tremendous similarity between them," as Dr. Hillyard puts it, they’re not identical. Anytime a new observation is made in the literature, and anytime a new drug becomes available, there will be differences in accommodating that information into the algorithms. "It is a problem," Dr. Hillyard says.

Reports typically include a citation of the resistance mutations that have been found by sequencing, as well as sensitivity/resistance or likely sensitivity/likely resistance. For a physician who truly understands the meaning of a particular mutation and is familiar with the current literature, that may be enough. "They can look and see there is a mutation that would change an amino acid at this codon position, for example," explains Dr. Hillyard. But such physicians are few and far between. At ARUP, he says, the HIV resistance testing laboratory engages a practicing clinician as its medical director, "because it is just so incredibly complicated now."

Reports also contain a summary and a prediction of whether a patient will be resistant or sensitive to specific drugs. This information comes from an algorithm of analyzing the sequence data, the rules of which are put together based on recommendations from a panel of experts. Those experts, says Dr. Hillyard, are treating physicians and others steeped in the latest research, who understand the implications of nucleic acid changes. Those changes alone reveal nothing-a genotype, after all, is completely dependent on correlation with in vitro phenotype studies and with actual clinical studies.

What reports don’t contain, says John Baxter, MD, is guidance on whether one drug is necessarily more appropriate than another drug, or what combination of drugs to use. Eventually that information may become available via a Web-based resource. The nonprofit HIV Resistance Response Database (hivrdi.org) is developing a computerized system that will offer suggested treatment regimens based on the mutations and treatment history provided by clinicians. "It’s sort of an artificial intelligence system, using neural networks to ’train’ this computer to be able to produce these best or likely most effective regimens," says Dr. Baxter, associate professor of medicine, Division of Infectious Diseases, Cooper Hospitals, University of Medicine & Dentistry of New Jersey.

The experts who update the algorithms may have significant differences of opinion regarding new data—how it should be interpreted and what weight to give it. "For something as complicated as HIV genotype testing, it’s probably impossible to have just a single algorithm that’s standardized for everybody," Dr. Hillyard says. Despite the fact that everyone is basically on the same, complicated page, "there are cut-out areas where there is vigorous debate what the significance of a given mutation is."

What these experts talk about—and for that matter, what any HIV specialist needs to be aware of—is that nothing stands still. HIV is like an endless series of switchbacks. They take you up, they take you back down, and sometimes they just make you sit down in the middle of the trail and cry.

New mutations can and do appear. So do new drugs. Then come new rounds of interactions and what they mean. "That’s why I won’t interpret genotypes myself," says Dr. Holland. "It’s more accurate and time-effective to use algorithms derived by a panel of experts."

The next thing to come down the pipe will likely be drugs that target something other than protease and reverse transcriptase—the very things sequenced by genotypes. These days there are also frequent conversations about so-called pattern recognition and which pattern best predicts lack of activity. Particularly within the nucleoside reverse transcriptase inhibitors, there’s a "split" between TAMs (thymidine analog mutations) versus non-TAMs. This is potentially valuable information that has yet to find a practical place in genotype reports. Replicative capacity is another hot topic—but only a phenotype, not a genotype, can provide insight into the virus’ fitness.

Dr. Baxter, who sits on the panel of experts (as does Dr. Wainberg) that updates the guidelines for Bayer’s TruGene system, talks about why their job is so hard. "You’ve got the pull of wanting the most accurate rules based on published literature on the one hand. On the other hand, information is rapidly being produced and presented, and you want to get that information into the rules as well, because it changes the way we understand resistance. So it’s a tug-of-war. And different algorithms may be accurate but out-of-date."

Sometimes the debate is whether to include new but unpublished information. "There may be a couple of abstracts, which suggests an important new finding. So that’s really a judgment call. We take it case by case," Dr. Baxter says.

Stirring things up a bit more, Dr. Reimer asks, "How good is the information we have to say that doing resistance testing makes any difference in how all patients will do?" Taking a poke at current thinking, he notes that a recent meta-analysis of genotype and phenotype studies hinted that in long-term patient outcome, the answer was, "Not much."

That’s not necessarily discouraging news, he says, because the studies under review were older, and interpretative success may be higher today. But he suggests it may be premature to make rock-hard pronouncements about the value of genotyping and phenotyping. "Most of the outcomes data that’s been presented is simply on viral load numbers; it’s not about disease progression, death, opportunistic infections, things like that," he says.

That brings him to his closing argument. "There’s a huge bandwagon for resistance testing: ’Oh, this is critically important, it’s standard of care, it ought to be done.’" And maybe it should, he agrees. "But I would say we should take a healthy, skeptical view of all of this. Because we really don’t know that it improves patient care. Physicians with a lot of experience, who are expert in HIV management, may do close to as well as testing does."

The problem is so few physicians fall into that category, says Dr. Hoover, who finds it frustrating when physicians fail to use genotyping information. Dr. Hoover was an author on the GART study, one of the first controlled trials of resistance testing. That trial showed a clear benefit, based on virologic response, of selecting therapy based on resistance assays. "I’ve talked to clinicians who just don’t use it," she says. "Some actually get the information as part of a clinical trial but have told me point-blank, ’We don’t use that for treating the patient.’

"I don’t understand it. I guess they’re confident that they know what’s going on," she says, obviously puzzled.

Richard Haubrich, MD, associate professor of medicine, University of California, San Diego, spends about 80 percent of his professional life designing, conducting, and analyzing clinical HIV trials. Here’s how he breaks down the evidence for using resistance tests:

First, some 20 to 30 different retrospective analyses have explored whether viral resistance, defined by either a genotype or phenotype, predicts reduced response (in terms of HIV RNA reduction) to a new antiretroviral regimen. "The answer, almost unequivocally, is ’Yes,’" Dr. Haubrich reports.

The second step was to look at whether resistance testing helps clinicians select a better regimen. Ten to 15 randomized studies looking at this have, in Dr. Hau brich’s opinion, yielded another explicit "Yes." Resistance testing, he says, does improve clinicians’ ability to select the next regimen. The endpoints of those studies are, again, better suppression of HIV, usually defined by viral load changes or proportion less than detectable, Dr. Haubrich notes.

There are other debates, of course, but most clinicians don’t care about the issues that get laboratorians talking—in this case, whether a system uses a slab gel or capillary electrophoresis method. Nor do they need to. "I think, as with most forms of testing, clinicians basically assume that the results are generated by a gradually improving and, hopefully, state-of-the-art method," Dr. Hillyard says.

For the record, Dr. Hillyard reports that in recent years there has been a migration toward capillary electrophoresis, accompanied by newer generations of reagents. Overall, he says, the quality of sequencing has gone up. And Dr. Reimer adds that genotype tests have become highly accurate in recent years. He recalls samples being sent to multiple labs four or so years ago, which revealed "a lot of discrepancy from one lab to another on whether they could detect mutations. I think that’s pretty much resolved now with the kits from the two companies."

What remains discrepant, from lab to lab, is the interpretation.

Some laboratories take the position that theirs is an in-house interpretation, based on in-house information that uses published data. And that’s fine, Dr. Hoover says. "But the docs need to understand that the same sequence may not provide the same interpretation, depending on the interpretation algorithm the testing laboratory is using."

It’s also worthwhile for laboratories—and probably their clinicians—to know which general method is being used. Anyone who reads the clinical trials and the primary literature will see that the studies are done using a particular reagent and accompanying algorithm, says Dr. Hillyard. "And it’s probably worth understanding what the laboratory’s experience has been in terms of what the limits of detection can be. Most laboratories can get down to 500 copies or so—but does that laboratory comfortably interrogate a low sample that has 500 copies?" Such information should be published in the laboratory’s user’s guide. Dr. Hillyard notes some clinicians may not appreciate the potential for sampling problems that arise near the cutoff levels. "You may get into trouble if you push the envelope down to the very, very lowest possible viral load levels with these methods," he says.

Dr. Holland, at William Beaumont Hospital, says she occasionally has difficulty explaining to referring clinicians that viral loads may simply be too low to obtain a useful genotype. In one case, she recalls, a clinician "argued with three techs before she got to me, because her patient had a viral load of 81, and we wouldn’t do the genotype." The previous viral load had been 79, and the clinician viewed the uptick as evidence of therapeutic failure. "I said, ’They’re not failing—they’re doing really, really well. I’m not going to be able to help you. It’s [the genotype] not going to be accurate, because we’re not going to be able to pick up minor species with that low of a viral load.’

"And then she got so mad at me she slammed the phone down," says Dr. Holland, who’s quick to add this response isn’t typical.

That blind spot is not exclusive to clinicians, she says. Before genotyping was brought onboard at Beaumont, she and her colleagues sent their samples to a reference lab, which, she says, would do a genotype on any sample regardless of how low the viral load was.

Laboratories no doubt appreciate this, but clinicians may not: Interpretation is not black-and-white. "You have to look at patterns, and to understand what their implications are," says Dr. Reimer. Clinicians also don’t always appreciate the basic biology of the virus and the mutability of its configuration, which can alter the sites of drug attachments, he says.

"Clinicians need to understand that, as a first principle, these tests absolutely do not stand alone," says Dr. Hillyard. It’s not a shortcut, allowing physicians to prescribe or change therapy based on a simple reading of the sensitivities/resistances. Clinicians need to consider what the current therapy is, what past therapies have been, and what prior resistance testing has shown, he continues. "There are, for a given patient, ways of estimating what likely drug resistance you may encounter. And you have to have thorough knowledge of drug interactions. So once you’re given this report, you’re just beginning to challenge your own thinking process as a clinician to use the data."

"The level of complexity of just doing this in the first place, then throwing in the idea that the different methods may not agree with each other, makes it very complicated for docs to try to interpret," says Dr. Reimer. Add to that the fact that there nothing intuitive about genotyping, though that may change as labs use gene analyses to look for resistance in other organisms. That’s already starting to happen with staph and tuberculosis, Dr. Reimer points out. But for now, the antibiotics/susceptibility worldview, which relies on phenotyping, holds sway.

Most clinicians probably do understand that these assays sequence a population. The commercial assays will pick up a resistant variant that’s present in perhaps 30 percent or more of the population of a patient’s virus. Finding archived specimens requires ultraspecialized techniques.

Another seemingly obvious but perhaps overlooked point is that any type of resistance testing needs to be done when the patient is adhering to therapy. Otherwise, the assay may not reflect the actual resistance of that patient’s virus—the wild type may re-emerge, giving the clinician false assurance that the patient will respond to specific medications.

Clinicians might not appreciate the laboratory’s own possible limitations in providing interpretations. Less-experienced labs may be supplying a software-based interpretation and nothing else, says Dr. Haubrich. "Unless the laboratory director is really interested in HIV resistance testing, he’s probably not going to add to that." In these cases, if clinicians don’t have a good idea of how to interpret the results, they should seek expert opinion. "Resistance testing is helpful, but it’s also important to know how to use the information," says Dr. Baxter.

In some centers, clinicians have relatively easy access to such experts, and can bring the difficult questions or cases to them directly. Occasionally they may call on the laboratory. There are also Web-based sources, such as Stanford University’s HIV drug resistance database (hivdb.stanford.edu), where clinicians can plug in mutations from a genotype report and be given an interpretation and comments.

There’s also the possibility that the need for expert opinion may eventually fade. While clinical trials have demonstrated such need, the technology was relatively new when those trials were done, so the expert advice would have added value that may not be needed in the future. But who knows for sure?

Hovering nearby is another related test: virtual phenotyping. Virtual phenotypes are, essentially, enhanced genotype reports. They aid interpretation of the genotype by finding close—not exact—matches to the patient’s genotype in a large database. The average phe no type for matched genotypes is reported.

"I haven’t seen clear studies that show it’s better than genotyping," says Dr. Hoover. "So my question is, why are we converting to something that doesn’t have a track record? I think a lot of it is because they [the clinicians] like the way it looks. Phenotype is something the docs are used to, because they do it for microbes.

"At least the genotyping has been through peer review, because there are two FDA-approved genotyping systems," she continues. She’s less sure about virtual phenotyping. "However, I’m not the clinician out there with the patient, and the virtual phenotype may be helping those clinicians make decisions. I guess that’s the difference between coming from the laboratory side, and the science side, versus the clinical side."

Given these chaotic discussions, it’s not surprising there are tremendous differences from practice to practice in how genotypes are ordered. But perhaps the greatest impediment has nothing to do with medicine or biology. "By far, the biggest reason why resistance testing is not done is cost," says Dr. Reimer. "Nothing else comes even close."

True, the test is reimbursable. But that doesn’t necessarily help those in public clinics that rely heavily on state or federal funding. "Those programs are really hurting these days," observes Dr. Reimer. Genotyping, an expensive test, becomes an obvious, if not ideal, target.

In California, reports Dr. Haubrich, money woes have forced physicians to make bleak choices. "We have caps on how many resistance tests we get, so in the clinic we sit there and say, ’OK, am I going to test all these naïve patients, or am I going to test someone who’s been on therapy and likely has resistance? It’s tough."

When cost doesn’t trip up clinicians, lack of self-confidence can, says Dr. Wainberg. "Some physicians won’t order a test if they’re not sure how to interpret it. It’s almost as though they’re intimidated, quite honestly," he says.

The solution, in his view, is to educate clinicians and patients alike about the importance of genotyping. Patients should insist on a genotype before therapy is started, in his view. And laboratories, he says, "should do a selling job to the clinicians they serve, in terms of always keeping them abreast as to the importance of requesting a genotype." Labs, he says, are in the ideal position to back up a request that more genotypes be done, with actual data that document the importance of genotypes and phenotypes.

The word "snapshot" comes up frequently in discussions about HIV genotypes. It’s an interesting analogy.

The earliest tools of HIV testing, CD4 counts and viral loads, coupled with clinical progression and outcome, were insufficient, recalls Dr. Hillyard. Thus, any new tool—in this case, genotyping—would have been anticipated as potentially having tremendous value, particularly since resistant viruses were seen as trouble early on. "Genotyping was anticipated very eagerly," says Dr. Hillyard.

Given that anticipation, a letdown was almost unavoidable. As genotyping made clinical inroads, the results have not had the same earth-shaking impact that viral load has had on monitoring. Studies looking at genotyping’s impact have been positive, but not overwhelmingly so, says Dr. Hillyard. Patients whose treatment plans used genotyping results had viral load measurements half a log lower than those without. That’s an important change—but hardly the all-or-nothing effect many hoped for.

Because genotyping is so complicated, and so difficult, it may never be the breakthrough so many have hoped for. It may remain a snapshot, and, like a snapshot, overlooked by some and valuable to others.

Karen Titus is CAP TODAY contributing editor and co-managing editor.