HIV Genotype and Phenotype—Arresting Resistance?

Fourteen antiretroviral drugs are now Food and Drug Administration (FDA)–approved for the treatment of human immunodeficiency virus (HIV) infection. Patients must take combination therapy perpetually because rapid virus turnover and high mutation rates promote drug resistance. The increasing prevalence of resistance has prompted development of drug resistance assays, and high-throughput techniques are now commercially available.

This issue of THE JOURNAL contains recommendations, developed by an international panel of experts, for the use and interpretation of anti-HIV drug resistance assays.¹ Two major analytic approaches are discussed: genotypic assays that identify particular mutations, usually point mutations, associated with resistance; and phenotypic assays that measure the ability of the patient's virus to grow in the presence of known concentrations of anti-HIV drugs. Although these techniques are scientifically sophisticated and exciting, they present dilemmas for the clinician and patient.

First, these assays are expensive. Despite automation, charges for a single genotypic assay are in the range of $400 to $550; for a phenotypic assay, $700 to $1000. Most third-party payers do not cover these assays. If an insurer denies payment, the patient is charged, or, in a large academic referral center, the center often bears the cost.

Turnaround time is slow for both assays—usually 2 to 4 weeks from the time blood is drawn until results reach the physician. This is too slow to have an impact in many clinical settings; information is needed quickly for managing occupational needlestick exposures and primary, or acute, HIV infection. But the problems of cost and time are overshadowed by the reliability of these tests.

Both genotypic and phenotypic assays have fundamental problems with sensitivity and specificity. Genotypic assays generally provide better sensitivity than phenotypic assays, with the capacity to detect a virus that constitutes as little as 10% of the patient's circulating virus population. But this depends on plasma HIV RNA concentration (viral load) at the time of sequencing and may miss some mutants that constitute up to 50% of the patient's viral pool.^{2 - 3} Because these assays amplify sequences from a mixture of HIV virus, only the predominant variants, or subspecies, within an individual are amplified.

Not surprisingly, academic laboratories frequently underestimate genotypic resistance.⁴ In large part, this reflects an "ongoing need for adequate standardization and clinical validation" of the assays.¹ Questions also exist about the relative difficulty in amplifying DNA from virus of different HIV-1 clades.

Both resistance assays require a plasma HIV concentration of 500 to 1000 RNA copies/mL to generate sufficient polymerase chain reaction product for analysis. Viral load assays used for monitoring patients now have sensitivities down to at least 50 copies/mL. Resistance assays are therefore not reliable in patients with viral loads of 50 to 1000 copies/mL, many of whom are now considered to have failed therapy.

Phenotypic assays also produce false-negative results; samples can contain a mixture of drug-sensitive and drug-resistant virus, thus underestimating resistance. Patients who have had treatment failure with multiple drugs may harbor low levels of drug-resistant virus in quiescent CD4⁺ lymphocytes, even if such mutant strains are not detectable in the plasma.^{5 - 6} Because a drug-resistant virus can grow out rapidly and become the dominant virus within a few weeks under pressure of drug treatment, it is important to identify these minority populations. Therefore, a negative resistance assay provides little reassurance to the clinician or patient.

To avoid false-negative results, the International AIDS Society–USA (IAS-USA) recommendations indicate that blood for resistance testing should be drawn while the patient is still taking a failing regimen.¹ Some physicians may find this impractical, even unethical.

Low specificity is also a concern with both types of assays. In the 2 commercially available phenotypic tests, resistance is defined as a 2.5- to 4-fold increase in the concentration of drug required to inhibit viral replication by 50% (IC₅₀).¹ By comparison, pneumococcal resistance to penicillin is defined as at least a 33-fold increase in the concentration of drug required to inhibit bacterial growth.⁷ If HIV drug resistance phenotyping standards were applied to penicillin, many patients would be denied a safe, effective, and inexpensive antibiotic. A more biologically rational definition of resistance incorporates the relationship between drug sensitivity and drug concentration in the patient.

Point mutations affect HIV drug sensitivity differently for different drug classes. A single point mutation may increase the IC₅₀ 1000-fold for nonnucleoside reverse transcriptase inhibitors (NNRTIs),⁸ while single point mutations generally increase the IC₅₀ for protease inhibitors only 3- to 5-fold.⁹ To use the penicillin analogy, if in vitro sensitivity changes by only a few fold, then the drug should still inhibit replication of the organism in the patient if drug concentrations are many fold higher than the IC₅₀ of mutants with low- or intermediate-level resistance.

False-positive results also can occur with genotypic drug resistance testing. Although 20% of patients in 1 report had baseline mutations associated with some degree of NNRTI resistance, initial response to efavirenz-containing regimens was as good as that in similar patients who lacked these mutations,¹⁰ thus calling into question the clinical relevance and predictive value of the test.

The complexity of interpreting genotypic results limits their value for the nonexpert. For example, some mutations, such as M184V in reverse transcriptase, reduce susceptibility to lamivudine but increase susceptibility to zidovudine. High-level resistance to protease inhibitors may require accumulation of 3 or more key mutations, and drug sensitivity may be affected by mutations outside the normal genetic "hot spots." Some gag and gag-pol mutations involving protein cleavage sites are outside the protease gene and can confer resistance to protease inhibitors.¹¹ Genotypic assays do not allow examination of such sequences. The 2 phenotypic assays amplify the protease and reverse transcriptase genes from virus in patient plasma and insert them into a recombinant vector system in which other HIV genes are derived from a standard laboratory strain, which misses resistance conferred by mutations outside the target genes. A "virtual" phenotype can be created by reporting the average drug sensitivity of virus containing a given set of mutations.¹² However, to be complete such analyses become complex: a gene as simple as that for HIV-1 protease, which contains only 99 amino acid codons, produces 4²⁹⁷ possible genotypes and nearly as many phenotypes.

Incorporating these assays into the management of HIV-infected patients may seem difficult. Several retrospective studies found an association between genotype or phenotype and virologic response to therapy, as described by Hirsch et al.¹ It is no surprise that drug resistance is a risk factor for treatment failure; the important question is whether access to genotype or phenotype data benefits the patient. To date, only 3 prospective controlled studies of the clinical utility of resistance testing have been reported,^{13 - 15} and, of these, just 1 was published in a peer-reviewed journal.¹³ None of these studies was blinded, and, for 1, only preliminary data were presented.¹⁵ The positive and negative predictive values of resistance testing were not calculated in these studies, although all reported better short-term virologic outcome with access to resistance data.

In the VIRADAPT Study,¹³ treatment-experienced patients for whom genotype data were available had a slightly better mean viral load decrease after 6 months (1.15 vs 0.67 log₁₀ copies/mL) and were about twice as likely to have a viral load less than 200 copies/mL (32% vs 14%) than patients who received standard care without genotyping. In the Genotypic Antiretroviral Resistance Testing (GART) Trial,¹⁴ patients were randomized to treatment recommendation by an expert panel with access to resistance data or to standard care. Short-term results were similar to those of the VIRADAPT study, with a modest improvement in mean viral load decline (1.17 vs 0.62 log₁₀ copies/mL) in the tested group after 8 weeks. However, since the test group was the recipient of regimen advice from an expert panel and the control group was not, it is possible that clinical benefit was the consequence of expert opinion rather than genotype data.

In preliminary analysis of outcomes in 127 treatment-experienced patients in a trial to examine the benefit of phenotypic resistance testing (VIRA3001), a greater mean viral load decrement at week 16 (0.87 vs 0.43 log₁₀ copies/mL by intention to treat) and more patients with a viral load of less than 400 copies/mL (62% vs 33%) were observed in the group with access to phenotypic data.¹⁵

However, the 2 genotyping trials (VIRADAPT and GART) were conducted in patients with baseline viral loads greater than 10,000 and 26,000 copies/mL, respectively. Whether these results will apply to patients with lower viral loads, or to treatment-naive subjects, is unknown.

In the IAS-USA recommendations published in this issue of THE JOURNAL resistance testing is advocated for all patients who have failed a treatment regimen and for pregnant HIV-infected women. The IAS-USA group also suggests consideration of resistance testing for patients with primary infection and prior to starting treatment in patients with established infection. This covers every HIV-infected patient except those who refuse therapy or are already fully suppressed on an established regimen. The firmest grounds for recommending resistance testing—as an adjunct for patients failing their first or subsequent multiple-drug regimen—are weakened by the fact that other important causes of treatment failure, such as nonadherence, intermittent inadequate drug concentrations, or both, may be difficult, if not impossible, to exclude with assurance.

To recommend testing all HIV-infected pregnant women may be premature because assessment of the mother's treatment history and treatment response may provide sufficient information to avoid initiating inactive drugs. Some drugs may be active in preventing vertical transmission of HIV even if the maternal HIV appears to be genotypically or phenotypically resistant. The fetus should receive drugs with a track record and known safety profile rather than untested antiretroviral combinations.

A somewhat different classification scheme distinguishes primary from secondary resistance. In primary resistance, the patient is infected by a drug-resistant virus variant. In secondary resistance, the patient acquires drug resistance mutations during therapy. These have very different treatment implications. Primary resistance for most approved antiretroviral drugs is uncommon at present but may be increasing in prevalence.^{16 - 17} It is in this setting that resistance testing is likely to have its greatest impact. Secondary resistance is a common cause of treatment failure, but studies and models suggest that this is most likely to be caused by noncompliance.¹⁸ Intervention for secondary resistance should focus more on improving adherence and regimen tolerance and less on picking effective agents from a computer printout.

Cost-effectiveness is also an important consideration. As the authors of the recommendations point out, resistance testing should not be used as "the principal criterion for deciding when to initiate or change therapy."¹ How much, then, should patients and third-party payers be willing to spend on such a diagnostic adjunct? If genotypic or phenotypic testing is ordered several times a year for the same patient, charges could exceed $3000. For some patients with secondary treatment failure, it seems that those same monies might be more effectively spent, for example, on directly observed therapy managed by a visiting home nurse.

Of note, many of the shortcomings of resistance testing outlined herein were pointed out in the HIV treatment guidelines published by the IAS-USA group in January 2000.¹⁹ The recommendations presented in this issue of THE JOURNAL endorse the value of resistance testing in several clinical settings despite being unable to approach the issues in a traditional evidence-based manner, a limitation alluded to by the authors themselves. An HIV-specialist panel assembled by the US Department of Health and Human Services, in guidelines published electronically at about the same time, also supports resistance testing.²⁰

To paraphrase the IAS-USA recommendations, when determining antiretroviral treatment, practitioners must consider treatment history, viral load, tolerance, adherence, and concomitant medications and disease.¹ However, despite the great promise of these technologies, it is still unclear how much resistance testing adds beyond the careful consideration of these other important factors. Only controlled, prospective studies can define the magnitude of benefit. Adequate evidence may never exist if resistance testing becomes standard practice now.