Furthermore, we are learning that newly infected patients who have never been on antiretroviral therapy are being infected with resistant forms of the HIV-1, which can result in suboptimal responses to their initial treatment regimens. According to recent estimates from a study sponsored by the Centers for Disease Control (CDC) across the U.S., about 15% of newly-diagnosed patients have drug resistance.
Antiretroviral (ARV) resistance testing is considered standard of care and is widely employed in the management of HIV-infected individuals. Current guidelines from expert panels recommend resistance testing in the setting of treatment failure and more recently have recommended that newly-infected individuals, those who have been infected for less than two years, have resistance testing prior to initiating antiretroviral therapy because of the increased prevalence and transmission of drug resistance.
Despite the widespread use of resistance testing in clinical practice, there remain a number of challenges to the clinician in applying these technologies to optimally manage the treatment of HIV infection, especially for the treatment-experienced patient.
When both genotypic and phenotypic tests are used in order to optimize treatment decisions, many times the results appear to be in conflict with one another -- where the genotype is interpreted to show resistance while the phenotype does not show resistance, or vice versa.
Moreover, mutational interactions can lead to phenotypic "hypersusceptibility" which appears to have clinical relevance.
Again, the resistance picture becomes more complex as patients stay on therapy longer and their virus accumulates an increasing number of resistance mutations.
And finally, measurements of viral replication capacity (or fitness) are now available to the clinician. This is a measure of how capable the virus is of reproducing. Some resistance mutations may allow the virus to multiply in the presence of drug, but the virus pays a price: it multiplies at a lower rate than the wild type virus. The role this in vitro (laboratory) measure should play in the management of patients remains to be fully defined.
JB is a 45-year-old interior designer who has been HIV-positive since 1989. He has been on a number of antiretroviral treatments over these years, starting with Retrovir (AZT) monotherapy and then Videx (ddI) monotherapy. His first PI was hard-gel saquinavir (Invirase) with Retrovir and Epivir (3TC). He was also treated with Sustiva (efavirenz), Zerit (d4T) and Videx (ddI) in the past.
For the past several years he has been maintained on Retrovir/Ziagen/Epivir and Kaletra (lopinavir with a ritonavir boost), but he is complaining of progressive lipoatrophy (loss of fat in the arms, legs, and sometimes in the face) and ongoing gastrointestinal (GI) distress. JB's viral load has never been completely suppressed, and generally runs between 300 copies to 5,000 copies. His CD4 counts remain in the mid-300's after reaching a low of 50 in 1994.
JB's doctor orders both a genotype and a phenotype as she considers changing the patient to a new regimen. The genotype shows a large number of resistance mutations to the nucleosides (NRTIs or nukes) and a mixture of K103N/K, related to resistance to the NNRTIs. In addition there are also many PI-related resistance mutations.
The phenotype, however, shows that JB's virus is still sensitive to Zerit, Videx, and Viread. It also shows sensitivity to the NNRTIs. The PIs show high-level resistance which is in agreement with the genotype.
JB's doctor is now wondering if she can use Sustiva and Viread as part of a new regimen. She is confused by the results in which the genotype interpretation suggests that both Viread and Sustiva would not be active while the phenotype report suggests these two drugs would be active.
For some time now it has been clear that patients who have initial virologic breakthrough (increased viral load) on an antiretroviral regimen do not necessarily have resistance to all of the drugs in that failing regimen. At first this did not make sense to many clinicians, who asked, "How can a regimen be failing, if not all the drugs in that regimen are failing?" The likely answer is that only one or two active drugs are often not sufficiently potent to maintain high levels of viral suppression and therefore the viral load comes back.
Drugs in the regimen that have a relatively "low genetic barrier to resistance" (where only a single mutation results in loss of antiviral activity), such as the nucleoside reverse transcriptase inhibitors (NRTIs) Epivir and Emtriva or the non-nucleoside reverse transcriptase inhibitors (NNRTIs), such as Sustiva and Viramune, will select for resistance rapidly in the early phase of virologic breakthrough.
Other drugs like Retrovir and the protease inhibitors (PIs) will have relatively slower evolution of drug resistance because of the requirement of accumulation of multiple mutations to confer high-level resistance. This means that clinicians may be able to recycle certain elements of a failing regimen if resistance testing is performed early enough in the course of virologic failure.
Boosting the protease inhibitor concentration using small doses of another PI such as Norvir appears to further protect the protease inhibitor component of the regimen from the early development of resistance in failing regimens. This was first demonstrated in a clinical trial where lopinavir boosted with Norvir (Kaletra) was compared to unboosted Viracept. The investigators showed that in the Viracept arm there was more PI resistance when virologic failure first showed up compared to the Kaletra arm.
This protective effect seemed to extend to the other drugs in the treatment regimen -- such that Epivir resistance was detectable in 29% of the Viracept arm compared to only 7% in the Kaletra arm after 96 weeks of follow up. This boosting effect is not unique to Kaletra -- it has been demonstrated for boosted Lexiva (fos-amprenavir) and more recently reported for boosted Invirase.
The first randomized controlled trial of resistance testing in the setting of treatment failure, the GART study, showed improved short-term control of HIV when treatment was guided by resistance testing compared to no resistance testing.
Since that time there has been controversy about the relative role of the resistance testing versus the expert advice that usually accompanies the resistance test results in improving outcomes. Patients who had resistance testing in the GART study also had the benefit of a resistance expert's opinion that accompanied the test result. So what was responsible for the improved outcomes, the expert's opinion, the resistance test, or the combination of the two?
The relative role of additional expert advice compared to practitioner-only genotype testing interpretation was clearly demonstrated in the HAVANA clinical trial. The investigators randomized patients on failing antiretroviral regimens to either receive genotype testing or not and either with or without expert advice. The group that had both genotyping results and expert advice had the best outcomes -- 69% of this group achieved a viral load of less than 400 copies/mL at 24 weeks. However, the group that had expert advice alone had outcomes that were comparable to the group that had genotyping alone, 49% compared to 46% achieving less than 400 copies/mL, respectively.
So given the complexity of interpreting resistance test results, it appears that having an expert help guide the treatment decisions improves treatment outcomes. However, one wonders how good the experts are at interpreting genotypes and how much agreement there is in the interpretation of genotypes among experts worldwide. We answered these questions in the GUESS study, in which we asked a panel of 12 international resistance experts to interpret 50 complex genotypes.
The experts had various levels of accuracy in predicting the phenotypic fold change based on the genotypic results for the 16 antiretrovirals commonly in use. For most drugs, the experts' accuracy was roughly 25-40%. The exceptions were the NNRTIs and Epivir, where levels of accuracy reached 75%. Levels of agreement between the experts were also around 40% for most of the drugs.
Despite these relatively low levels of agreement, the expert panel agreed on the treatment recommendations about 80% of the time. So one can conclude that experts are not terribly accurate in translating genotypes into phenotypes or expected drug activity levels, but there is broad agreement in making treatment recommendations based on the genotype -- which in the end is what the average clinician seeks from a genotype interpretation algorithm.
The bottom line is which drugs should be used in the setting of resistance and on that the experts seem to agree pretty well.
Stanford University: On the Stanford University web site at http://hivdb.stanford.edu/ is a database, HIVdb. This is an expert system where users provide genotypic sequences and are provided with levels of resistance to approved anti-HIV drugs. For each drug, each resistance mutation is assigned a drug penalty score. The total score for a drug is the total of the scores of each mutation present that is associated with resistance to that drug. Using the total drug score, the program reports one of the following levels of inferred drug resistance: susceptible, potential low-level resistance, low-level resistance, intermediate resistance, and high-level resistance.
Aptivus: The recently approved protease inhibitor (PI) Aptivus (tipranavir) was developed specifically to deal with HIV that already has protease resistance mutations. But which ones and how many? How can a clinician know whether it's worth trying Aptivus? The resistance score that the drug's developer, Boehringer Ingelheim, has developed demonstrates how complex these analyses are getting to be.
Researchers analyzed the results of Phase II and Phase III trials and identified 21 mutations at 16 different codons that reduce susceptibility of HIV to Aptivus or reduce HIV's response to Aptivus. About half of these mutations are unique for Aptivus, such that HIV often remains sensitive to Aptivus even when many PI mutations that came about from older PIs are present. The total number of these mutations generates the resistance score to Aptivus. Boehringer Ingelheim has determined the score at which Aptivus starts to lose its effectiveness, and a higher score where it seems to have no effect.
But will these complicated resistance scores ever be calculated and used in general practice? Will they be reflected in commercial genotypic tests? Or will the testing companies and HIV clinicians rely on simpler ways to predict whether or not to use a particular drug? There are two simpler approaches for Aptivus that will probably be used more frequently. The first is to count the number of "key" protease mutations present in a patient's virus. The key mutations seen to affect Aptivus are at codons 33, 82, 84, and 90. A virus with two or fewer is sensitive to Aptivus; one with three of these mutations has decreased sensitivity; and one with all four is resistant. The other simple method is to obtain a phenotype. A baseline phenotypic test that shows a fold change below 3-fold indicates that HIV is still sensitive to Aptivus and is associated with a good treatment response defined as a 1 log (10-fold) or greater drop in viral load after six months of treatment. A fold change between 3 and 10 results in decreased sensitivity to Aptivus and somewhat less of a benefit. Finally, a fold change above 10 indicates resistance to Aptivus.
As mentioned earlier (see "Phenotypic Test"), the interpretation of the phenotype comes in defining the "cutoffs" for drugs. The ARV drug concentration required to inhibit a patient's virus strain compared to a reference wild type virus strain without ARV drug resistance is normally expressed as a fold change in IC50 (the drug level or inhibitory concentration that inhibits 50% of virus growth), where the virus becomes less susceptible to a given drug.
There are two potentially important cutoffs for each drug. The lower cutoff defines when the susceptibility begins to decline but the drug still has partial activity. The upper cutoff would be the fold change where all drug activity is lost.
These so called "clinical cutoffs" have not yet been defined for most of the antiretrovirals in use today. Therefore the clinician depends on what are called "biological cutoffs." These are based on the variation of fold changes seen in viral populations without drug resistance. If a patient's sample is outside that "normal" distribution, then it is considered to have reduced susceptibility.
The number of drugs with clinical cutoffs is few because of the challenges inherent in defining these cutoffs. For most drugs, resistance and response is on a continuum and therefore any cutoff is by its very nature somewhat arbitrarily drawn.
Secondly, most drugs are used in combination with other drugs, so teasing out the impact of a particular drug is difficult. Nonetheless progress is being made here as well. The PhenoSense assay by Monogram Biosciences and the Virco Antivirogram assays now report clinical cutoffs for several antiretrovirals. Having clinical cutoffs should improve the clinical utility of phenotype testing -- particularly in experienced patients with complex genotype patterns.
Since both phenotype and genotype are imperfect tests, many clinicians -- like the one in our case example -- order both tests to try to bring all information to bear when making the critical decision about a patient's new antiretroviral regimen. Although there are no clinical trials that support the use of both genotyping and phenotyping together to improve outcome, this approach seems to have clinical merit, particularly in the setting of highly treatment experienced patients with multi-drug resistant HIV-1 infection.
The problem with using both tests is discordant results -- that is, the genotype appears to say one thing and the phenotype the opposite, as our case study illustrates. So the risk of more information is more confusion rather than clarity.
Discordance occurs commonly when both tests are ordered, where the results can be discordant as much as 25% -30% of the time. The causes of discordance can include mixtures of wild type and mutant virus; virus that has "back-mutated" to an intermediate form that doesn't affect the phenotype but can quickly revert to a drug-resistant version if the drug is re-started; and interactions between mutations that can appear to cancel each other out in terms of the phenotype.
Comparison of Genotyping and Phenotyping
|Genotypic Testing||Rapid results (1-2 weeks)||Indirect measure of resistance|
|Less expensive than phenotyping||Some mutations have questionable impact on resistance|
|Mutations may show up before phenotypic resistance||Minority viral species are not detected|
|Widely available||Mutational patterns can be difficult to interpret, especially for treatment-experienced patients|
|Better than phenotype at detecting mixtures of resistant and wild type virus||Genotypic interpretation may not be relevant for non-B subtypes of HIV|
|Phenotypic testing||A direct and quantitative measure of resistance||Susceptibility cutoffs vary for each testing company|
|Can be applied to any antiretroviral, including new drugs for which resistance mutations haven't been identified||Clinical cutoffs have not been defined for some agents; the level of fold change that matters may not be clear|
|Interactions among mutations are reflected in test results||Minority viral species are not detected|
|Accurate with HIV subtypes other than subtype B||Long turnaround (3-4 weeks)|
|More expensive than genotyping|
Hypersusceptibility can be seen as the opposite of resistance or reduced susceptibility. Here the patient's virus requires less concentration of drug to inhibit growth compared to the control wild type virus. This laboratory phenomenon has been shown to have clinical relevance, most clearly for the NNRTI class. There have been several cohort studies and clinical trials that demonstrate that patients with HIV that is hypersusceptible to the NNRTIs have a better short-term virologic response.
It has been shown that NNRTI hypersusceptibility is related to the presence of NRTI-associated resistance mutations, but the relationship is a complex one and difficult to predict from a genotype. Therefore, although it is not clear how a clinician could incorporate hypersusceptibility into a treatment plan, the phenotype has the advantage as a direct measure of this phenomenon compared to a genotype.
Beyond phenotype and genotype tests, clinicians are now also faced with the question of how to incorporate measures of viral fitness into their treatment strategies for HIV-infected patients. This is particularly true for treatment-experienced patients in whom a fully suppressible regimen is not possible. These patients may consistently have a low viral load with stable, but low, CD4 counts.
Viral fitness refers to the ability of the virus to multiply in a patient. It includes the effect of antiretrovirals, immune suppression and other factors. Although we do not have a true viral fitness measurement, replication capacity (RC) -- a laboratory measurement -- has been shown to correlate with measures of fitness in patients.
One study showed that resistant virus had a lower replication capacity than the wild type virus, which could explain why it was "overgrown" fairly quickly by the wild type virus. Replicative capacity may develop as a tool for predicting patient outcomes that is independent of viral load testing and resistance testing.
JB, the patient in our case study (and his physician) are fortunate to be at a hospital where there is research going on regarding viral resistance. The physician was able to get an expert on resistance to review the patient's test results. The expert's recommendation was that Viread should be considered since the patient had never taken it. However, he didn't trust the phenotypic result showing sensitivity to Sustiva -- JB had taken it previously -- and recommended against using it.
Another piece of good luck for JB is that one of the research studies being done at the hospital is a trial of a new NNRTI drug that is supposed to be active against HIV that already has NNRTI resistance. His doctor talked to the study coordinator and got JB enrolled in the study. However, she still has to consider how many "active" drugs the patient will have. Because of accumulated resistance, even a "new" drug may not be active against the virus. So with Viread and the experimental NNRTI, she still wants another active drug.
Videx and Zerit both showed up as possibilities on the phenotype, but JB has taken both of them before. Resistance might be archived and might come back out quickly. Also, Zerit has been linked to lipoatrophy, which JB is complaining about. The doctor discusses this situation with JB and he agrees to try Fuzeon (enfuvirtide) even though it requires injections twice a day. Since it attacks a totally different part of HIV's life cycle than any drug JB has taken, it will almost certainly be an active drug for him.
As part of their discussion, JB's doctor told him that in several recent clinical trials of new drugs, the patients who added Fuzeon along with an experimental drug did better than others in the trial who did not use Fuzeon. This approach will also let JB get off of Kaletra, which might be contributing to his stomach problems -- and, since it is still working and there is no sign of resistance to it, JB will still have the possibility of going back on it if the experimental drug doesn't pan out.
The results were very good for JB. He tolerated the twice-daily injections of Fuzeon, and the experimental drug worked well. He was very careful about not missing any doses of his new regimen, and within three months, his viral load was undetectable (less than 50 copies/ml) for the first time in years. His CD4 count had not gone up too much, now about 400, but JB and his doctor were hopeful that it will continue to climb.
Online Resistance Resources
Antiviral Mutations at www.mediscover.net/antiviralintro.cfm
Mediscover Infectious Diseases is provided as a resource center for researchers and clinicians that specializes in the study and management of infectious disease.
HIV Resistance Response Database at www.hivrdi.org/
The aim of the HIV Resistance Response Database Initiative is to improve the clinical management of HIV infection by developing and making freely accessible a large clinical database and bioinformatic techniques that define with increased precision and reliability the relationships between HIV resistance and virologic response to treatment.
HIV Resistance Web at www.hivresistanceweb.com/
HIVresistanceWeb is an independent, educational resource dedicated to the advancement of anti-HIV therapy through information sharing and expert discussion of current issues in antiretroviral drug resistance and clinical HIV virology.
HIV Sequence Database at http://hiv-web.lanl.gov/content/hiv-db/mainpage.html
(From Los Alamos National Laboratories.) The HIV Sequence Database focuses on five primary goals:
• Collecting HIV and SIV sequence data (since 1987)
• Curating and annotating this data, and making it available to the scientific community
• Computer analysis of HIV and related sequences
• Production of software for the analysis of (sequence) data
• Publication of the data and analyses on this site and in a yearly printed publication, the HIV Sequence Compendium, which is available free of charge.
Stanford University HIV Drug Resistance Database at http://hivdb.stanford.edu/
See the site home page for information on the resistance tools available here.
Companies offering resistance test products
Monogram Biosciences at www.monogrambio.com/200HIVProducts.aspx
Virco Lab at www.vircolab.com
The management of HIV infection is getting increasingly complex and specialized as more drugs become available, patients are treated for longer periods of time, and more complex tests become available to manage patients and ARVs. Resistance tests are now standard of care and a clinician must have a solid understanding of the tests, including their limitations, to be considered qualified to manage HIV-positive individuals. Although the technology behind resistance testing is reliable and fairly standardized at this point, the interpretation of resistance testing is still evolving. There are many online resources available for clinicians to consult and that patients should know about as well. The Stanford University database is one such resource that provides comprehensive and up to date information about drug resistance.
Test results need to be interpreted with a good understanding of their limitations, and how to assess conflicts between genotypes and phenotypes. Incorporating new tests like replicative capacity to optimize management will continue to challenge even the most expert clinician. In the end, all tests need to be interpreted in the clinical context of the patient, which includes assessment of the treatment history, viral load and CD4 profile, adherence patterns, and toxicities and other adverse events. It is only in this broader context that optimal use of resistance tests can be achieved.
Andrew Zolopa graduated from UCLA in 1984 and completed his residency as chief resident in internal medicine at Santa Clara Valley Medical Center in 1988. After serving as a Robert Wood Johnson Clinical Scholar and completing a postdoctoral fellowship in infectious diseases at Stanford in 1992, Zolopa joined the faculty at UCSF School of Medicine in the Department of Epidemiology and Biostatics. In 1994 he returned to Stanford as director of the Positive Care Clinic and joined the Stanford faculty as assistant professor of medicine, Division of Infectious Diseases and Geographic Medicine. He is an active researcher and teacher in addition to his clinical practice and leadership.