LABORATORY TESTS FOR MONITORING HIV-1 INFECTION

Daniel R. Kuritzkes, MD
Associate Professor of Medicine and Microbiology
University of Colorado Health Sciences Center
Co-Director, Colorado AIDS Clinical Trials Unit

Clinical significance of drug resistance in HIV-1 infection

Treatment of HIV-1 infection with potent combination therapy (sometimes called A highly active antiretroviral therapy, or HAART) dramatically slows progression to AIDS. In the developed world, the wide-spread use of HAART has resulted in a sharp decrease in the number of deaths due to AIDS over the last five years (1). Although transient improvements in survival have been demonstrated with nucleoside monotherapy and dual combination therapy, the clinical benefit of such regimens may be limited by the eventual development of drug resistance. The high error rate of reverse transcriptase (the viral enzyme responsible for reproducing the viral genome, or genetic material), coupled with high levels of HIV-1 replication result in the very rapid emergence of drug-resistant strains of HIV-1 in most settings where treatment fails to completely suppress virus replication.

Current triple therapy regimens are able to delay the development of drug resistance because they suppress virus replication to undetectable levels. However, problems with adherence to treatment, drug toxicities, differences in drug absorption or metabolism (i.e., pharmacokinetics), and other host factors can compromise the activity of a HAART regimen. Over time, these factors may allow the accumulation of mutations that confer drug resistance, leading eventually to treatment failure. Although individual drugs select for specific resistance mutations, the rate at which these mutations emerge is quite variable, and often difficult to predict. In particular, failure of a triple-drug regimen may be associated with resistance to only one or two drugs in the regimen (2). Therefore, knowledge of the specific pattern of drug resistance may be helpful in choosing the next treatment regimen. For this reason, considerable effort has been devoted to the development of assays for HIV-1 drug resistance. As a result, the tests have improved considerably over the last few years and are quickly becoming an essential tool in choosing therapy for patients experiencing treatment failure.

Genotype and phenotype

Resistance testing generally involves determining the genotype or phenotype of a virus. The genotype refers to the sequence of nucleotides in the viral genome that determine the genes of the virus. Genotypic assays are assays that determine the nucleotide sequence of specific genes, or parts of genes. Viruses that have the same gene sequence as other viruses found in nature (for example, from patients who never received antiretroviral therapy) are considered to be Awild type@. Genetic differences (mutations) that arise during the course of antiretroviral therapy are considered to be drug resistance mutations if the presence of these mutations reduces the susceptibility of the virus to a particular drug. In some cases the nucleotide sequence at a particular position in a gene may vary from one isolate to the next, even in the absence of any drug treatment. Such differences are often referred to as polymorphisms.

The phenotype refers to the characteristics or properties of the virus. Phenotypic assays for drug susceptibility determine the amount of drug needed to inhibit viral growth in tissue culture. The amount of drug needed to inhibit virus growth by 50% is called the 50% inhibitory concentration, or IC50; similarly, the concentration of drug that inhibits virus growth by 95% is known as the IC95. Testing a particular drug against a large number of isolates from patients who never received antiretroviral therapy can determine average IC50 for wild-type isolates of HIV-1. Viruses that are inhibited by similar concentrations of that drug are considered susceptible or sensitive; those that are inhibited only at higher drug concentrations are considered resistant.

Results of phenotypic assays are sometimes expressed as Afold-resistance@ by comparing the IC50 of the patient=s virus to that of a control isolate. For example, if the IC50 for zidovudine of the control isolate is 2 nM and the patient isolate has an IC50 of 20 nM, then the patient=s virus would be 10-fold resistant as compared to the control. However, the definition of Aresistant@ also needs to consider the concentration of drug that can be achieved in the plasma and the relationship between IC50 or fold-resistance and clinical response to the drug in question.

How resistance tests are done

Genotypic assays

Several approaches to genotyping are available, ranging from full-length sequencing of the target gene to point mutation assays, which focus only on a particular mutation of interest. The most commonly used genotypic assays rely on automated DNA sequencing. Using this technique, the nucleotide sequence of some or all of the gene of interest (e.g., protease [PR] or reverse transcriptase [RT]) is obtained, then translated into the predicted amino acid sequence in order to determine whether specific mutations are present or absent. Automated sequencing offers the most complete data on viral genotype, but generates more information than is needed for most clinical purposes. For example, HIV-1 RT has 550 amino acids, but mutations at only a small number of these positions are implicated in drug resistance. Therefore, interpretation of the genotype is needed in order to help distinguish which changes are merely polymorphisms and which might be significantly associated with drug resistance.

Most commercially available genotypic tests rely on automated sequencing technology. Viral RNA is extracted from a sample of plasma and reverse transcribed into complementary DNA in the laboratory. The PR- and RT-coding regions of the cDNA are then amplified by polymerase chain reaction (PCR), and the nucleotide sequence of the PCR product (or amplicon) is determined on an automated DNA sequencer. Some laboratories use specific kits developed by companies such as ABI/Perkin Elmer or Visible Genetics, Inc. to perform genotyping. Usually the kits provide standardized reagents needed for the RT-PCR and DNA sequencing steps. Other laboratories use so-called "home brew" assays using reagents and primers developed individually by each laboratory. A list of the mutations most often associated with resistance to currently available drugs is given in Table 1.

Other types of genotypic resistance assays such as the Line Probe Assay (LiPA; Innogenetics) (3) or the differential probe hybridization assay being developed by Bayer (4) are designed to provide more limited information by testing for the presence or absence of specific mutations at particular positions, or codons. These assays have the advantage of being faster and less complex than standard genotyping, and may be more sensitive at detecting minor species. However, because these tests do not generate a comprehensive sequence, information needed to interpret complex genotypes might be missing. Furthermore, the tests must be reconfigured to include important new mutations as they are defined.

Phenotypic assays

In the past, phenotypic assays required the isolation and preparation of high-titer stocks of virus from patient samples. This procedure was laborious and time-consuming, requiring 6 to 8 weeks to generate a result. Because primary clinical isolates grow best in peripheral blood mononuclear cells (PBMC), these assays made use of activated PBMC from seronegative donors. However, PBMC from different donors vary in their ability to support the growth of HIV-1, leading to significant interassay variation (5). For this reason, several groups have developed recombinant virus assays, in which the viral genes of interest (e.g., PR and RT) are introduced into a plasmid that carries all of the other viral genes needed for replication in cell culture (6; 7).

As with genotypic assays, the first step in a phenotypic assay involves extraction of HIV-1 RNA from plasma, followed by reverse transcription and PCR amplification of the PR and RT genes. These amplified genes are then inserted into vectors used to generate recombinant viruses that can be tested for susceptibility to protease and RT inhibitors. Because the viruses are each identical except for the protease and RT sequences, most of the interassay variation observed in older PBMC-based assays is eliminated, allowing for very precise determination of the fold-resistance of a particular viral isolate. As a result, differences in susceptibility of 2.5- to 4-fold compared to control usually are considered to be significant in the currently available assays.

Phenotypic assays are more complex and labor intensive than genotypic assays. Automation of these assays makes it possible to test many samples simultaneously, and allows for high through-put. However, the complexity of the automation limits the availability of these assays to only a few laboratories. at present phenotyping by recombinant virus assay is available from two sources: Virco (AntiVirogram; Mechelen, Belgium) and ViroLogic (PhenoSense; South San Francisco, CA). It is unlikely that these assays will ever be produced as "kits" that can be used to perform phenotypic tests at local hospital laboratories.

Genotyping vs phenotyping: advantages and disadvantages

Genotypic and phenotypic assays each have specific advantages and disadvantages. Because many laboratories are capable of performing automated DNA sequencing, genotypic testing is available from many more laboratories than is phenotyping. As a rule, genotyping is less complex, faster, and less expensive than phenotyping. In addition, in some circumstances a key resistance mutation may be detected by genotyping even though no change in the phenotype is produced (e.g., emergence of an L90M mutation in protease). Such changes might be the first step along the path to high-level resistance, and detection of these mutations might prompt a change in therapy in a patient with detectable plasma viremia.

Despite these advantages, there are important disadvantages to genotyping. Perhaps the greatest disadvantage is the complexity of the data generated by these assays. Keeping track of which mutations correspond to resistance to which drug poses an enormous challenge. Moreover, mutations that cause resistance to one drug might improve viral sensitivity to different drug (for example, the 184V mutation causes resistance to 3TC but sensitizes HIV-1 to AZT (8)). Another potential disadvantage of genotyping is inter-laboratory variation. An international study compared the performance of laboratories on a blinded panel of samples containing wild-type or mutant viruses in different proportions (9). Nearly all the labs provided the correct sequence for specimens that were completely wild-type. However, mutations in RT were detected only 66% of the time, and mutations in protease were detected only 71% of the time in samples that contained only mutant virus. The expected mutations were identified correctly in fewer than half the laboratories when mutant and wild-type virus were present as a 50:50 mixture. Results of this study showed that the experience of the technician actually doing the assay was critical to laboratory performance.

Phenotyping has the advantage of providing susceptibility data in a format familiar to most clinicians (i.e., the IC50 or IC90). In addition, because susceptibility is measured directly, the effects of mutational interactions are more easily sorted out. As a result, there is less need for expert interpretation of phenotypic resistance tests. Because these assays are performed in a few central reference laboratories, interlaboratory variation is less problematic. A study conducted by the CDC compared results on paired plasma samples tested by the Antivirogram and PhenoSense assays (10). The two assays gave similar results in over 90% of samples tested. Most of the samples on which the labs disagreed had phenotypes that were close to the cut-offs (e.g., 2.5- or 4-fold resistance). Further comparisons of a larger number of resistant isolates are planned.

Phenotypic assays also have potential disadvantages. Because these tests are less widely available, there is often a longer time until results are available. In addition, phenotypic assays are significantly more expensive than genotypic assays. Another limitation of phenotypic assays is that clinically significant "cut-offs" or "break points" for distinguishing sensitive and resistant isolates have not been defined for most drugs (see below). Moreover, phenotypic tests may fail to detect very small shifts in susceptibility that accompany the presence of only one or two key resistance mutations that nevertheless result in reduced drug activity (e.g., the 90M mutation for saquinavir or the 74V mutation for ddI).

Shared limitations

Genotyping and phenotyping share some common technical limitations as well. Samples with <500-1000 HIV-1 RNA copies/mL usually fail to yield a result. In addition, both types of tests are relatively insensitive to the presence of minor variants. As a rule, a mutant needs to make up at least 20-30% of the viral quasispecies before it is detected by genotyping or can exert a noticeable effect on the phenotype. Because genotyping and phenotyping both rely on a PCR step to amplify PR and RT gene sequences, cross-contamination is a significant concern. Even though diagnostic laboratories take elaborate precautions to prevent contamination, problems can occur even in the best laboratories. Therefore, if a resistance test result does not seem to make sense in the context of a patient's current or former treatment regimen the test should be repeated.

Interpreting resistance tests

Various systems for interpreting HIV-1 genotypes have been developed. Most interpretations use a "rules-based" approach. That is, a group of experts determine which mutations or combination of mutations are associated with resistance to specific drugs, and establish an algorithm for interpreting the genotype. These algorithms are used as the basis of automated computer-generated reports, as in the TruGene assay. Such algorithms require periodic updating as new information becomes available. Alternatively, clinicians may refer to one of a number of on-line databases, some of which now offer genotype interpretation (Table 2). Genotype reports from diagnostic laboratories usually come with some form of rules-based interpretation indicating to which drugs the virus has developed resistance.

Although results of phenotypic assays seem to be straightforward, interpreting phenotypes in fact may be more complicated than originally believed. For example, an important issue is how to define "cut-offs" for susceptible and resistant viruses. Both the PhenoSense and AntiVirogram assays use laboratory strains of HIV-1 as controls. Until recently, cut-offs were defined solely on the basis of interassay variation. For example, if the IC50 of wild-type control viruses varied by 4-fold, then viruses with IC50's that were more than 4-fold greater than the control were considered to be resistant; those with IC50's that were less than 4-fold above the control were considered to be susceptible. This approach has two short-comings: first, the natural variation in susceptibility of wild-type isolates might be greater than the interassay variation for the laboratory control strain, and second, these cut-offs were not based on clinical response data.

These shortcomings have been addressed by modifications in reporting for both the AntiVirogram and PhenoSense assays. Virco has defined new ranges for "sensitive" and "resistant" based on the distribution of susceptibility of isolates from 1000 treatment-naïve patients (11). This change has narrowed the range of IC50 values that fall within the sensitive category for the nucleoside RT inhibitors, but considerably widened the range of IC50's included under the susceptible category for NNRTIs. ViroLogic has re-defined resistance to d4T and ddI as a >1.7-fold increase in IC50 for these drugs as compared to control based on data that suggest a reduced response to these drugs in viruses with only modest reductions in susceptibility. Similarly, clinical response data resulted in redefinition of resistance to abacavir as an IC50 >4.5-fold above control (12).

Even the concept of "cut-offs" presents certain difficulties, however. For example, the cutoff for lopinavir has been defined as a >10-fold increase in IC50 above control. However, clinical trials show that there is a continuous relationship between lopinavir susceptibility and virologic response (13). Two-thirds of patients with virus that had 20- to 40-fold resistance to lopinavir still achieved virologic suppression on lopinavir-containing regimens. Data suggest that even NRTI's may have residual activity against partially resistant viruses (14). Thus, it might be more realistic to consider drugs as having greater or lesser activity against viruses that are more or less susceptible to the particular drugs.

The Virtual Phenotype

A third approach to resistance testing is the "virtual" phenotype. This assay, developed by Virco, is really a genotype resistance that is interpreted with the aid of a large database of samples with paired genotypic and phenotypic data. Viruses with genotypes that are similar to the patient's virus are identified by searching the database, and the average IC50 of these matching viruses is calculated. This information is then used to estimate the likely phenotype of the patient's virus. Since the database currently has >18,000 paired genotypes and phenotypes, there is a high likelihood that a large number of matches can be found for most genotypes encountered in practice. The actual and virtual phenotype show excellent correlation (r2>0.8) for most drugs (15). Moreover, retrospective analysis of samples from the VIRA 3001 study (see below) showed that the VirualPhenotype performed as well as the actual phenotype in predicting response to treatment (16). Several studies are underway to test the performance of the VirtualPhenotype in patient management.

The VirtualPhenotype has important strengths and weaknesses. The main strength of this approach is that it reduces complex genotypic data to simple phenotypic categories based on a rational, data-driven analysis of similar genotypes. Reports indicate the number of matches, and the proportion of sensitive and resistant viruses among the matches. It is important to remember, however, that the virtual phenotype only provides an estimate of the probable phenotype of the patient's virus-the actual phenotype could be more or less sensitive than the average value obtained from the database. The main weakness of this approach is that the confidence placed in the result depends on the number of matches, and on picking the right codons to incorporate into the database search. Correlation between actual and virtual phenotype will be weaker for newer drugs or in cases where there are fewer matches due to unusual genotypes. Also, the virtual phenotype may place undue emphasis on using genotypes to predict phenotype, rather than treatment outcome.

Prognostic value of resistance testing

Retrospective analyses of many studies have established the value of resistance testing as a predictor of treatment response in patients changing therapy. In these studies, resistance testing was performed on stored samples from patients experiencing virologic failure that had been obtained at the time a new regimen was initiated. Results of the resistance assays were used to classify patient viruses as sensitive or resistant to the drugs in the new regimen. In a meta-analysis that combined results of several retrospective studies, resistance to the drugs in the new regimen predicted a significantly greater likelihood of treatment failure (17). Conversely, the more drugs in the new regimen to which the virus was sensitive (as predicted by the resistance assays), the lower the chance of treatment failure. Several of these studies showed that drug resistance remained a significant independent predictor of the likelihood of treatment failure even after controlling for treatment history. In other words, resistance testing provided provided significant additional prognostic information over and above the information available from treatment history alone.

Prospective trials of drug resistance testing

Viradapt. The Viradapt study was a randomized trial of genotyping versus standard of care to guide the choice of salvage regimen in patients failing antiretroviral therapy (plasma HIV-1 RNA levels >10,000 copies/mL after at least 3 months of a PI-containing triple-therapy regimen) (18). Patients in the genotyping arm had a significantly greater reduction in plasma HIV-1 RNA at 6 months as compared to patients in the control arm. Similarly, patients in the genotyping arm were more likely than patients in the control arm to achieve a plasma HIV-1 RNA levels <200 copies/mL at 6 months. Follow-up of these patients indicated that these benefits persisted out to week 48 for patients in the genotyping arm. Further analysis of a subset of patients in this study demonstrated the importance of achieving adequate plasma levels of PI's for optimal treatment response, even after taking into account the benefits of genotypic drug resistance testing (19).

GART. A multicenter, randomized pilot study to determine the short-term effects of genotypic analysis in management of patients failing antiretroviral therapy provides additional supportive evidence (20). In this trial (CPCRA 046, or "GART" [Genotypic Antiretroviral Resistance Testing]), 153 patients with an increasing plasma HIV-1 RNA level after at least 16 weeks of treatment with a triple-drug combination regimen were randomized to receive either genotypic data and expert interpretation or standard clinical management. At week 12, patients in the genotyping arm had a significantly greater reduction in plasma HIV-1 RNA from baseline as compared to those in the control arm. The proportion of patients achieving a plasma HIV-1 RNA <500 copies/mL was also significantly greater in the genotyping arm. Within both treatment groups, the virologic response correlated with the number of active drugs prescribed.

Several caveats should be kept in mind regarding the GART study: 1) "expert" advice regarding patient management was provided only to the genotyping group; 2) follow-up to date is limited to 12 weeks; 3) expert advice was ignored by the treating physician in a substantial proportion of the genotyping patients. Nevertheless, these results are consistent with the conclusions of Viradapt.

HAVANA. A study that addresses directly the role of expert advice is the Havana study, conducted in Spain. This randomized study compared the utility of genotypic resistance testing, expert advice, or both as compared to standard of care in selecting salvage regimens for patients experiencing failure of antiretroviral therapy (plasma HIV-1 RNA level >1000 copies/mL). Two hundred forty-seven patients were enrolled. Genotyping and expert advice each resulted in significantly better virologic response. The best response rates were observed in patients who received both genotyping and expert advice. These results suggest that although expert advice is helpful, the availability of genotypic resistance assays leads to further improvements in virologic outcome of salvage therapy (21).

VIRA3001. The VIRA3001 study tested the clinical utility of phenotypic resistance testing in selecting a salvage regimen (22). In this study, patients failing their first PI-containing regimen were randomized to phenotyping (using the Antivirogram assay) or standard of care. Two-hundred twenty-one patients were enrolled. Patients in the phenotyping arm had a significantly greater reduction in plasma HIV-1 RNA level by week 16, as compared to patients in the standard of care arm. Using an intention-to-treat analysis in which missing patients were considered as failures, the proportion of patients achieving the primary endpoint (HIV-1 RNA 400 copies/mL) was not significantly different in the two arms. However, this difference was significant in a secondary analysis limited to actually observed values (59% vs 42%; P=0.033). Similar increases in CD4 cell counts were observed in both arms. Of note, very few patients entering this trial had prior NNRTI experience. Overall, patients in the phenotyping arm received significantly more new drugs to which the virus was sensitive than was true for the control arm.

NARVAL. French investigators compared genotype (TruGene), phenotype, and standard of care in a randomized trial of 541 patients failing a 3-drug PI-containing regimen with plasma HIV-1 RNA levels >1000 copies/mL (23). Most patients were heavily pre-treated, having received a median of seven drugs. An important difference between this study and other studies is that in the phenotyping arm, investigators provided clinicians with a ranked list of possible regimens based on the fold-resistance resistance testing of the drugs; no such listing was offered to patients in the genotype or standard of care arms. Expert advice was followed by clinicians approximately 85% of the time. Patients were stratified at entry on the basis of virus load and prior treatment experience. No significant difference between arms was found at week 12 for either the percent of patients with plasma HIV-1 RNA <200 copies/mL or for the percent of patients showing a >1-log10 decrease in virus load from baseline. There was a trend favoring the genotyping arm at week 24. Significantly more patients in the genotyping arm than in the standard of care arm achieved plasma HIV-1 RNA levels <200 copies/mL at weeks 12 and 24.

A number of factors may explain the failure to find a significant benefit of resistance testing in NARVAL. Patients in this study had substantially greater prior treatment experience than in the GART and Viradapt studies. In addition, there were considerable differences between the phenotype and genotype arms in classification of patient samples with regard to resistance to d4T, 3TC, and abacavir (24). For example, 56% were considered resistant to d4T by genotype, but only 24% were resistant by phenotype. By contrast, 62% were considered 3TC-resistant by genotype compared to 81% by phenotype. In the case of abacavir, 84% were classified as resistant by genotype vs 41% by phenotype. This discrepancies could have led to incorrect treatment recommendations that may have altered the outcome of the trial. Despite these shortcomings, NARVAL does point out the importance of validating interpretation of genotypes and phenotypes on the basis of clinical outcome, and suggests that resistance testing may be less useful in highly treatment-experienced patients with few remaining treatment options.

The inhibitory quotient (IQ)

An alternative approach to interpreting resistance tests is to relate the observed IC50 of an isolate to the achievable trough concentration (Cmin) of drug or to the measured trough concentration of that drug in a particular patient. The relationship between drug exposure and drug susceptibility can be explored through the inhibitory quotient (IQ), which is the Cmin/IC50 ratio. (Many drugs are bound to proteins in the plasma, reducing the effective [free] drug concentration. Therefore, IC50 values used in calculating the IQ are often adjusted for protein binding, but how to make such adjustments is controversial.) A high IQ means that the trough plasma concentration is significantly greater than the amount of drug needed to inhibit the virus in question; a low IQ suggests inadequate drug levels or a highly resistant virus.

One study calculated IQ's for patients receiving ritonavir-boosted indinavir (IDV) therapy for treatment of IDV-resistant virus (25). Phenotypes were predicted from genotypic data using the Virtual Phenotype. Using the Virtual Phenotype and measured trough concentrations of IDV to calculate a "virtual" IQ, the investigators found that response rates were significantly higher among patients with an IQ>2 as compared to those with an IQ<2 (P<0.003). These results suggest that combining phenotypic data with drug levels might be particularly useful in predicting treatment response. However, adjusting drug doses on the basis of the IQ in an attempt to overcome drug resistance may not be advisable, since the safety of very high drug levels that might be required in certain cases has not been evaluated.

When and how to use resistance testing

Taken as a whole, the available data provide a compelling rationale for the use of resistance testing in managing antiretroviral therapy (26). Resistance testing is recommended for patients failing a current regimen in order to aid in the selection of salvage therapy. Resistance testing should also be performed in patients with primary (acute) HIV-1 infection. This recommendation is based on the observation that the prevalence of drug resistance in virus samples from treatment-naïve patients with recently acquired HIV-1 infection is approximately 10% (27). Another use of resistance testing is to optimize therapy in HIV-1-infected pregnant women.

Whether or not to perform resistance testing in all patients prior to initiating antiretroviral therapy remains controversial. Because wild-type revertants are likely emerge and outgrow less fit drug-resistant variants over time, resistance testing may not be helpful in guiding therapy for treatment-naive patients with established HIV-1 infection of more than 6-12 months' duration.

For the same reason, possible resistance testing in patients experiencing treatment failure should be performed while the patient is still receiving the failing regimen, whenever possible. Once a regimen is stopped, there is the possibility that residual wild-type virus will rapidly overgrow the less fit drug-resistant mutants, giving a potentially misleading test result. Thus, results of resistance testing are most reliable for the drugs the patient is taking at the time the test is performed.

How should the results of resistance testing be interpreted? Tables of mutations that confer resistance to currently available drugs are provided in most recent reviews of drug resistance, including guidelines for use of drug resistance testing recommended by the International AIDS Society-USA (26). A number of web sites may also be helpful in interpreting resistance test results (Table 2). In the case of phenotypic testing, an increase in the IC50 of 10-fold or more (as compared to control isolates) is clear evidence for resistance in the case of most drugs. For certain drugs, however, increases in the IC50 as small as 1.7- to 4-fold may provide evidence of significant resistance. If resistance to a drug is identified, then that drug is likely to have little or no activity and should not be used in a salvage regimen. Similarly, if resistance to a drug has ever been identified, it is safe to assume that resistant virus persists, even if not detected in a current sample. (Remember, plasma virus populations change rapidly when drugs are started or stopped, but resistant viruses selected by prior regimens persist in latently infected resting CD4+ lymphocytes and can re-emerge promptly if these drugs are restarted).

Ultimately, the best choice of therapy for an individual patient should be determined by taking into account all of the information available, including history, disease stage, virus load, CD4 count, and patient preferences. Because industry-wide standards for proficiency testing and quality assurance are still evolving, clinicians should not hesitate to question the results of resistance tests that seem to be at odds with the treatment history of a given patient. It is important to remember that resistance testing is only a tool, and is never a substitute for sound clinical judgement.

DIAGNOSING HIV INFECTION

A diagnosis of infection with HIV can be made by several different kinds of assays, include virus culture, PCR, and antibody testing. In most cases HIV infection is diagnosed by demonstrating the presence of specific antibodies in the blood. On average, antibodies against HIV become detectable in patients with acute HIV-1 infection approximately 25 days after infection; by 12 weeks nearly all infected individuals are HIV antibody positive (28; 29). (The process of developing antibodies to a virus is termed seroconversion, and individuals who become antibody-positive are sometimes called seroconverters.) Testing for HIV antibodies is a two-stage process: sera that give a positive reaction by an initial screening assay are retested to exclude the possibility of clerical or laboratory error; repeatedly reactive sera are then tested by a confirmatory assay to verify that reactive antibodies are directed against HIV antigens. Assays have been developed for detection of HIV antibodies in serum, whole blood, saliva, urine, and dried blood collected on filter paper.

HIV ELISAs and western blots. Antibodies to HIV are usually detected by assays known as enzyme-linked immunosorbent assays (ELISA or EIA). The wells of plastic microtiter plates are coated with recombinant viral proteins (antigens), which stick to the plastic. If serum from an infected patient is added to the well, HIV antibodies bind to the proteins and become attached to the plate. After washing away the serum the bound antibodies are detected by a second antibody that is linked to an enzyme such as alkaline phosphatase. This second antibody binds to human anti-HIV antibodies and can be detected by reacting the plates with a substrate for alkaline phosphatase that turns color when cleaved by the enzyme.

HIV antibody tests are more than 99% sensitive (that is, they detect the presence of HIV antibodies in nearly all infected patients); virtually the only infected patients who are not detected by standard HIV tests are those who are tested within the first few weeks after infection. The specificity of HIV antibody tests is also better than 99%, but some false positive tests do occur. Therefore, all positive ELISAs are confirmed by a second test such as the western blot. Though not as sensitive as the ELISA, the western blot is more specific and allows the laboratory to identify the specific HIV proteins to which the antibodies are reacting. Only tests that are positive by ELISA and by a second, confirmatory test are reported as HIV positive.

HIV tests performed in the United States detect antibodies to both HIV-1 and HIV-2. Antibodies to most of subtypes of HIV-1 group M are detected by the current tests, but they are less reliable at detecting infection with more distantly related strains of HIV-1 belonging to groups O and N. (Group M accounts for the vast majority of HIV-1 infections; groups O and N are found predominantly in western Africa.)

Rapid HIV tests. In addition to the standard HIV ELISA, rapid diagnostic tests have been developed. Rapid diagnostic tests can be helpful in identifying HIV infection in women in labor who have not had pre-natal HIV testing (so as to identify those in need of antiretroviral therapy to prevent mother-to-child transmission of HIV) and in testing source patients in needle-stick injuries (so as to identify health care workers in need of post-exposure prophylaxis). The single-use diagnostic system (SUDS) HIV-1 test (Abbott Laboratories) has been the only FDA-approved test available for rapid diagnosis of HIV-1 infection in the United States. This 10-minute agglutination test uses HIV-1 antigen-coated latex beads, which are incubated together with patient serum or plasma to generate a result. Sensitivity and specificity are greater than 99% (30). However, at the present time the SUDS test has temporarily been withdrawn from the market due to manufacturing problems.

Home testing for HIV infection. Home sample collection kits for HIV testing are approved by the FDA for direct marketing to consumers in the United States (Home Access HIV-1 Test System, Home Access Health Corporation, Hoffman Estates, IL). The kit provides materials necessary to perform a finger stick in order to obtain a dried blood spot, which is then mailed directly to a central laboratory for analysis. The specimen is identified by an code number, which preserves the anonymity of the user. Results and counseling are provided to the user by calling a toll-free telephone number. These kits only for home specimen collection, not for actual self-testing at home, which is not FDA-approved. Some internet sites advertise kits for true home testing, but those tests have been found to be unreliable, and the FDA has issued warnings against their use.

Virus culture
HIV-1 can be cultured from plasma or peripheral blood mononuclear cells (PBMC) of infected individuals. A positive culture provides direct evidence of HIV-1 infection, but virus culture is rarely necessary to establish a diagnosis. Virus culture is used almost exclusively for research purposes, particularly with regard to viral growth properties and drug resistance.

Virus is easily cultured from PBMC of infected patients who are not on antiretroviral therapy. Cultures are performed by mixing patient cells with cells from uninfected donors that have been stimulated with phytohemagglutinin (PHA) and IL-2. HIV-1 can also be recovered from latently infected PBMC of patients with undetectable plasma viremia. Resting cells that carry latent virus usually belong the memory class of CD4 T cells and constitute an important reservoir of HIV-1 (31-33). CD8 T lymphocytes that are present in PBMC exert an antiviral effect and may prevent outgrowth of virus in vitro. Therefore, it is usually necessary to remove these cells in vitro in order to recover virus from patients on effective antiretroviral therapy.

p24 antigen assays
An alternative approach to diagnosing HIV-1 infection is to detect the presence of viral antigens in the blood. The best antigen for this purpose is the capsid antigen, p24, a viral structural protein that makes up most of the virus core particle. Because high titers of p24 antigen are present in the serum of acutely infected individuals during the short period between infection and seroconversion, p24 antigen assays are useful in the diagnosis of primary HIV-1 infection. After seroconversion the antigen is bound by p24-specific antibodies and becomes undetectable in the majority of infected individuals. For this reason p24 antigen assays are not useful for diagnosing HIV-1 infection in otherwise healthy individuals who are thought to have established infection.

Later in the course of disease, serum p24 antigen again becomes detectable in 30-70% of patients (34; 35). Presence of detectable p24 antigen is associated with an increased risk of clinical progression (36). Early studies of antiretroviral therapy used quantitative p24 assays to assess the antiviral activity of new drugs, but this assay has been replaced by virus load testing using RT-PCR or bDNA tests.

PCR assays

Qualitative assays for proviral HIV-1 DNA. Although only a small fraction of peripheral blood mononuclear cells (PBMC) from infected individuals carry proviral HIV-1 DNA, they can usually be detected by PCR. Therefore, a diagnosis of HIV-1 infection can be made by demonstrating the presence of proviral DNA in PBMC. Assays for detecting proviral DNA employ the polymerase chain reaction (PCR) to amplify conserved sequences in the HIV-1 gag or pol gene. Experienced laboratories can achieve 100% sensitivity and specificity in PCR testing for HIV-1 DNA (37). The sensitivity of HIV-1 DNA PCR assays in clinical practice is only 96-99%, however (38-41). Strict attention to guard against contamination from the carry over of PCR products is essential to prevent false-positive results.

As with virus culture and p24 antigen detection, sensitivity is lower in individuals with higher CD4+ cell counts due to the lower titer of circulating infected PBMC. DNA PCR assays for HIV-1 are used most often in the early diagnosis of HIV-1 infection in neonates. Clinical applications of these tests are relatively limited in adults, but occasionally DNA PCR testing may be helpful in resolving indeterminate western blots in high-risk individuals.

VIRUS LOAD TESTING

Measurement of plasma HIV-1 RNA levels (virus load) can be used to monitor the course of disease and the response to antiretroviral therapy in patients with HIV-1-infection. Assays based on different methods for quantifying plasma HIV-1 RNA assay have been developed. These include reverse transcription followed by polymerase chain reaction (RT-PCR) (Amplicor HIV-1 Monitor, Roche Diagnostic Systems), nucleic acid sequence-based amplification (NASBA; HIV-1 RNA QT, Organon-Teknika), and nucleic acid hybridization and branched DNA (bDNA) signal amplification (Quantiplex HIV-1 RNA, Bayer Nucleic Acid Diagnostics). A fourth assay, based on DNA hybridization and colorimetric detection (Digene assay; Digene Diagnostics) has also been developed, but is not widely available at this time.

In PCR-based assays, HIV RNA is converted into DNA by reverse transcription followed by PCR amplification of the DNA. The PCR product is detected by hybridization with an enzyme-conjugated probe specific for HIV-1, and quantified by reacting bound probe with a substrate that undergoes a color change, as in an ELISA. The branched DNA assay uses non-enzymatic means to amplify the signal from HIV RNA. In this assay, viral RNA is "captured" by hybridization to complementary oligonucleotides that are bound to the wells of a microtiter plate. The captured viral RNA target is then hybridized to branched oligonucleotides (hence the name "branched" DNA assay), which in turn are hybridized to enzyme-conjugated oligonucleotides that can be quantified as above. The NASBA assay is similar in concept to the RT-PCR assay except that reactions occur at one temperature. At present the Roche Amplicor HIV-1 Monitor (RT-PCR) assay is the only one approved by the U.S. Food and Drug Administration, and is the most widely used in clinical practice.

Results of the three commercially available quantitative HIV-1 RNA assays are highly correlated (42; 43). All three assays have a lower limit of quantification of approximately 50-80 copies/mL (Table 3). (Although the lower limit of the standard Amplicor HIV-1 Monitor assay is 400 copies/mL, the range of the assay can be extended by pelleting virion particles prior to RNA extraction, a modification commonly referred to as the "ultrasensitive" HIV-1 Monitor assay.) These assays are much less precise at plasma HIV-1 RNA titers below 200 copies/mL (44). Serial testing of clinically stable patients not on antiretroviral therapy (or on a stable failing regimen) has shown the relative stability of plasma HIV-1 RNA levels over the short term (weeks to months), with a biological variation of approximately 0.3-0.4 log10 copies/mL (45; 46). Given these factors, changes of greater than 0.5-0.7 log10 (3- to 5-fold) are likely to reflect significant changes in HIV-1 replication (47).

Although most strains of HIV-1 that circulate in North America belong to subtype B, more than 10 different subtypes are found around the world. The HIV-1 Monitor 1.0 (RT-PCR) assay is significantly less sensitive for detecting HIV-1 from subtypes A, E, and F as compared to the Quantiplex version 3.0 (bDNA) assay (48). Plasma HIV-1 RNA levels that appear to be lower than expected in a patient with advanced disease can be a clue to infection with a non-subtype B strain. Incorporation of alternative primer sets in the new version of the HIV-1 Monitor assay (version 1.5) has improved the ability of this assay to diverse HIV-1 subtypes.

Clinical utility of plasma HIV-1 RNA monitoring. Numerous studies have demonstrated the correlation of plasma HIV-1 RNA levels with stage of disease. Patients with AIDS or symptomatic HIV infection have significantly higher titers of plasma HIV-1 RNA than do those with asymptomatic infection. In addition, patients with higher virus loads are likely to progress more rapidly than patients with lower virus loads. For example, individuals with plasma HIV-1 RNA levels >100,000 copies/mL within six months of seroconversion are 10-times more likely to progress to AIDS within five years than patients with lower levels of plasma HIV-1 RNA (49). Plasma HIV-1 RNA levels are correlated with the rate of CD4 count decline and with the rates of progression to AIDS and death in untreated patients with established HIV-1 infection (50) (Table 4).

Most studies suggest that plasma HIV-1 RNA levels provide prognostic information even in late stages of disease (51). Similar results have been observed in children with perinatally acquired HIV-1 infection (52; 53). However, some studies suggest that the CD4 count is a better predictor of disease progression than is plasma HIV-1 RNA in patients with very low CD4 counts (below 50 cells/mm3) (54).

The rapid change in plasma HIV-1 RNA levels in response to treatment makes it possible to assess the effectiveness of antiviral therapy within a matter of weeks. The relationship between change in virus load and treatment benefit has been analyzed in several large clinical trials (55). These studies show that a decrease in plasma HIV-1 RNA confers a significant reduction in risk of disease progression, independent of baseline plasma HIV-1 RNA level and CD4 count, and independent of the increase in CD4 count due to treatment (56). Much of the benefit of antiretroviral therapy can be attributed to its effect on plasma HIV-1 RNA levels. A 0.3-log10 (2-fold) reduction in plasma HIV-1 RNA levels confers a 30% reduction in the risk of progression to AIDS or death (57); a 1-log10 (10-fold) reduction reduces the risk of disease progression by approximately two-thirds (58). Although initial studies suggested that HIV-1 RNA was a stronger predictor of response to antiretroviral therapy than the change in CD4 count, subsequent studies make clear the prognostic importance of improvement in both markers (59; 60).

Sample collection. Blood for plasma HIV-1 RNA testing should be collected into tubes containing EDTA as an anticoagulant, and the plasma separated and stored frozen at -70o C until testing. Studies show that HIV-1 RNA is stable for up to 48 hours at room temperature in the presence of EDTA, but ideally samples should be processed within 6 hours after collection (61; 62). Events leading to immune activation such as vaccination or acute infectious illness can transiently raise the plasma HIV-1 RNA level (63; 64). Therefore, plasma HIV-1 RNA testing should not be performed within four weeks of an intercurrent infection or immunization. Because of differences between assay formats and commercial laboratories, the same laboratory should be used for serial tests on an individual patient.

Current treatment guidelines recommend obtaining two measurements of plasma HIV-1 RNA to determine the baseline or "set-point" virus load (65). Virus load testing should be performed immediately prior to initiating treatment and repeated within 2-8 weeks of starting treatment in order to assess the initial response to a regimen. A 1.0-log10 decline in plasma HIV-1 RNA level is expected for treatment-naïve patients within 8 weeks of starting an initial antiretroviral regimen, and plasma virus should fall to undetectable levels (below 50 copies/mL) by 16 weeks. However, more than 24 weeks may be required for plasma virus titers to fall below the limit of detection in patients with high pre-treatment levels of viremia (above 100,000 copies/mL). Declines of 0.5-log10 or more within 8 weeks should be expected following a change in regimen due to treatment failure. Subsequently, plasma HIV-1 RNA levels should be repeated every 3-4 months in order to monitor the success of antiretroviral therapy.

IMMUNOLOGICAL TESTING

Although considerable attention in recent years has focussed on laboratory tests for measuring virus load and drug resistance, immunologic tests remain an essential part of patient monitoring. Ultimately, it is the loss of CD4+ T-lymphocytes that results in immune deficiency in HIV infection, and the ultimate goal of antiretroviral therapy is immune reconstitution. In clinical practice the CD4 count is the most commonly used marker of immune competence. However, the CD4 count is only a surrogate for immune function. A number of additional assays provide a more complete assessment of immune function. Although these assays are not routinely available for clinical purposes, familiarity with these assays is helpful in understanding much of the research regarding HIV pathogenesis and immune reconstitution.

CD4 counts

The CD4 molecule is expressed on the surface of helper T-lymphocytes. CD4 interacts with HLA class II molecules on the surface of antigen-presenting cells to help stabilize the interaction between the antigen-specific T-cell receptor on the T-helper cell and the antigen-HLA class II complex on the antigen presenting cell. T-helper cell function can be significantly impaired by blocking the CD4-HLA class II interaction. CD4 also serves as the primary receptor for HIV-1 and HIV-2. It is the specific affinity of the HIV envelope glycoprotein (gp120) for CD4 that targets HIV to helper T cells and macrophages.

Even prior to the identification of HIV as the cause of AIDS, the progressive loss of T helper cells was noted to be a characteristic finding in patients with this disease. On average, there is a loss of 30-60 CD4+ cells per year, although in many patients, CD4+ T-lymphocyte counts may remain stable for several years followed by a period of rapid decline (66; 67). Natural history studies and clinical trials have demonstrated that the CD4+ lymphocyte count is an independent risk factor for progression to AIDS and death. The CD4 count provides an estimate of the immunologic status of the patient, and therefore, is an excellent marker of the immediate risk of opportunistic infection. Such complications are rare in patients with CD4 counts above 500 cells/mm3. As the CD4 count drops below 500 cells/mm3 patients may begin to experience recurrent minor infections such as herpes simplex virus or oral candidiasis. The risk of more serious opportunistic infections increases significantly as the CD4 count falls below 200 cells/mm3.

The CD4 count increases promptly in response to antiretroviral therapy. The rise in CD4+ cell count is related to the extent to which virus replication is suppressed. Even patients who do not achieve complete virologic suppression may show significant increases in CD4 counts. Cohort studies have shown that patients who achieve a significant CD4+ cell increase in response to potent antiretroviral therapy have a substantially lower risk of disease progression, whereas those patients who achieve viral suppression but do not have an increase in their CD4 counts remain at increased risk of developing an AIDS-related opportunistic infection (68). Thus, monitoring the CD4+ cell count is an essential component of patient evaluation.

The CD4 count is determined by "staining" patient blood cells with antibodies to various cell surface markers, including CD4. The antibodies are conjugated to fluorescent tags that emit light of a certain frequency when excited by a laser beam. In flow cytometry the stained cells are fed in a stream past the laser, and the proportion of cells that emit light at the right wavelength is determined. The number of lymphocytes circulating in the blood (determined by a blood count) is multiplied by the percent of cells staining positive for CD4 in order to calculate the number of helper T cells.

CD4 counts are subject to considerable inter-assay variation. Most of this variation is due to fluctuations in the total lymphocyte count, rather than any inherent inaccuracy of the flow cytometry portion of the test. CD4 counts are subject to diurnal variation (that is, counts are higher in the morning than in the evening), and may change as a result of an acute illness. The percentage of lymphocytes that are CD4+ (the "% CD4") is less variable than the absolute CD4 count. The % CD4 cells is comparable to the absolute count in predicting the risk of disease progression or in assessing the response to treatment (66).

Guidelines suggest that CD4 + T cell counts be measured at the time of diagnosis and
generally every 3-6 months thereafter (65). CD4 counts can show considerable day-to-day variation. For this reason, any large unexpected change in the CD4 count should be confirmed by repeat testing a few days apart. The CD4 count trajectory over several follow-up visits may be particularly helpful.

Other flow cytometry markers

Numerous additional parameters can be measured by flow cytometry but their role in patient management is not defined. CD8+ T lymphocytes are cytotoxic T cells (CTL), which play an important role in protection against viral infection. The number of circulating CD8 cells increases significantly during primary HIV infection, and remains moderately elevated until very late stages of disease. However, most studies show that the CD8 count is not a significant independent marker of disease progression.

CD4 and CD8 lymphocytes also can be characterized according to their activation state by staining for CD38, HLA DR, or CD25. As with CD8+ cells, the proportion of activated cells (either CD4+ or CD8+) tends to be higher in infected patients with HIV infection as compared to control. The proportion of activated cells drops with initiation of effective antiretroviral therapy. Analysis of data from the Multicenter AIDS Cohort Study (MACS) have shown that the proportion of CD8+CD38+ cells is an independent prognostic marker of disease progression even after controlling for CD4 count and virus load (69). However, measurement of CD38 has not been adopted widely in clinical practice. Likewise, measurement of the proportion of memory and naïve CD4 or CD8 cells is of interest from the point of view of pathogenesis studies, but is not used clinically at this time.

Functional assays

Although CD4 cell counts are good predictors of the risk of opportunistic infection, they are an indirect measure of immune function. Several assays have been developed for testing specific immune responses to HIV-1 and to opportunistic pathogens such as cytomegalovirus (CMV). However, these assays remain research tools and have not been validated for clinical use.

Proliferation assays

Binding of antigen by T cells through the T-cell receptor triggers activation and proliferation. This response is antigen-specific, and requires the participation of appropriate antigen-presenting cells such as macrophages. For example, if PBMC from an individual who has received a tetanus shot are cultured together with tetanus antigen, the small number of tetanus-specific cells will begin dividing and secreting cytokines. Since the number of cells that proliferate is too small to detect by counting the cells, proliferation is usually detected by measuring the incorporation of 3H-thymidine (thymidine is incorporated into the DNA of dividing cells). The amount of 3H-thymidine incorporated in response to antigen stimulation is compared to incorporation in control wells to yield a stimulation index (SI). In general, SI's greater than 3 are considered evidence of an antigen-specific response. Antigen-specific responses can also be detected by assaying production of specific cytokines such as interleukin 2 (IL-2), IL-4, interferon-gamma, etc.

HIV-specific proliferative responses can be demonstrated in patients with acute HIV infection (70) but are quickly lost as a result of ongoing virus replication and depletion of HIV-specific CD4 cells. Proliferation in response to antigens from opportunistic pathogens such as candida, PCP, MAC, CMV, etc can also be demonstrated. These responses are lost as HIV disease progresses, and are gradually restored in patients after initiation of potent antiretroviral therapy (71; 72). However, the correlation between pathogen-specific levels of proliferation in vitro and the risk of disease in vivo has not been established. Cells can also be stimulated to proliferate non-specifically by complex plant carbohydrates (lectins) such as phytohemagglutinin (PHA). PHA-dependent proliferation is usually preserved until late stages of HIV disease.

Cytotoxic T lymphocyte assays

Cytotoxic T lymphocytes (CTL) function to destroy cells that express foreign antigens. CTL play an important role in the elimination of virally infected cells. The majority of virus-specific CTL are CD8+. In contrast to CD4+ helper cells, which recognize antigen presented in the context of class II major histocompatibility complex (MHC) molecules, CTL recognize antigens in the context of class I MHC molecules. The importance of this distinction is that most of the time antigens need to be produced inside the cell in order to be presented by class I molecules. By contrast, antigens presented by class II MHC molecules can be taken up by antigen presenting cells, processed, and then presented to CD4+ cells.

The classical assay for CTL activity is the chromium release assay (73). Target cells expressing HIV antigen on their surface are labeled with a radioactive isotope of chromium (51Cr). Patient cells are then mixed with the target cell and incubated for several hours. Lysis of antigen-expressing cells releases 51Cr into the medium. HIV-specific lysis is calculated by comparing lysis of target cells expressing HIV or control antigens in the presence or absence of patient effector cells, and is usually expressed as the % HIV-specific lysis. Several studies have shown an inverse correlation between virus load and HIV-specific CTL activity, suggesting CTL play a role in controlling virus replication. This view is supported by data from SIV-infected macaques in which CD8 cells were transiently depleted, resulting in a prompt rebound in virus load.

Instead of measuring cytotoxicity, the CD8+ CTL response can be assessed by measuring IFN-a production by HIV-specific effector cells in an ELISPOT assay. In this assay, antigen-presenting cells (APC) are immobilized on the plastic surface of a microtiter well, and effector cells are added at various effector:target ratios. The binding of APC's by antigen-specific effector cells triggers the production of cytokines including IFN-a by the effector cells. The cells can be stained to detect the presence of intracellular IFN-a and the number of positively staining foci (spots) counted under a microscope. A second method for quantifying the number of circulating antigen-specific CD8+ T cells is the tetramer assay. In this assay, a specific epitope is bound to synthetic tetrameric forms of fluorescently labeled MCH Class I molecules. Since CD8+ T cells recognize antigen in the form of short peptides bound to Class I molecules, cells with the appropriate T cell receptor will bind to the labeled tetramers and can be quantified by flow cytometry. Although this method is less time-consuming than the ELISPOT assay, the tetramer assay measures only binding, not function. Not all cells that bind a particular antigen necessarily become activated. However, a recent paper demonstrated good correlation between ELISPOT, tetramer, and cytotoxicity assays (74). Another limitation of the tetratmer assay is that tetramers of all Class I subtypes are not available, so that

Lastly, CD8+ T cells also produce a soluble factor that capable of blocking HIV-1 infection, known as CD8 antiviral factor (CAF) (75). The precise nature of this factor is not known, and it might represent a mixture of several molecules. Production of this factor is triggered upon binding of specific antigens to CD8+ cells via the T cell receptor, but can also be triggered non-specifically. Production of the factor is assayed by testing dilutions of culture supernatant from appropriately stimulated cells for the ability to inhibit HIV replication in vitro. Chemokines are elaborated by stimulated CD8+ cells, and share certain similarities with CAF, but studies indicate that CAF is distinct from RANTES, MIP-1a, and MIP-1b.

Although the various assays for assessing virus-specific CTL have improved over the last five years, they are performed only in research laboratories and are not validated for clinical use. However, study of HIV-specific CTLs in various stages of disease provides important insights into AIDS pathogenesis and ultimately may lead to development of effective vaccine strategies.

Therapeutic Drug Level Monitoring

Over the last few years there has been increasing interest in individualizing the dosing of antiretroviral therapy. Although careful dosing on the basis of weight or body surface area is standard practice in pediatrics, dosing in adults has tended to follow a "one size fits all" model. Doses generally are based on average pharmacokinetic profiles in volunteers, and dose ranging studies in HIV-infected subjects. Usually determining the best dose of a drug requires balancing antiviral activity and potential toxicities. However, there can be considerable inter-individual variation in drug absorption and elimination (76-78) Measuring plasma drug concentrations in patients could, in theory, identify patients receiving too little or too much of a drug, allowing patient-specific adjustments in dose to be made. Although the concept of therapeutic drug level monitoring (TDM) has gained in popularity over the last few years it remains controversial.

In the early years of antiretroviral therapy, little attention was paid to drug levels outside the pharmacology community because nucleosides have short plasma half-lives and because accurate measurement of active drug-intracellular nucleoside triphosphate (e.g., 3TC-triphosphate)-was available only in specialized laboratories. Nevertheless, one study did suggest that adjusting ZDV doses based on plasma drug levels resulted in optimum viral response to therapy (79). From practical purposes, however, TDM is best applied to drugs for which there is a clear dose-effect relationship, for which simple, precise, and reliable plasma assays exist, and for which there is a defined therapeutic range. The protease inhibitors and NNRTI's come closest to meeting these criteria. The relationship between trough plasma concentrations (Cmin) or area under the curve (AUC) and reduction in plasma HIV-1 RNA levels has been established for the PIs. Recent data regarding efavirenz (EFV) demonstrate a dose-relationship for antiviral activity and for central nervous system side-effects (80).

One misconception is that TDM can be a helpful adjunct to monitoring medication adherence. Because of the relatively short half-life of the NRTI's and PI's, random drug levels may provide some information regarding the most recent dose, but will not be helpful in determining longer-term adherence patterns. For example, a patient who frequently missed doses of a drug might be more likely to take his or her medication the day before a scheduled clinic visit, resulting in a plasma level that was "in range". Obviously, such results can produce a misleading picture of actual adherence.

A more appropriate use of TDM would be to ensure that adequate plasma drug levels are achieved in patients who are taking drugs with pharmacokinetic interactions (e.g., efavirenz and amprenavir), or to ensure that plasma trough levels exceed the drug concentration required to inhibit the patient's virus. As discussed in the section on drug resistance testing, the ratio between the plasma trough concentration (Cmin) and IC50 of the patient's isolate for a particular drug can be expressed as the inhibitory quotient (IQ). In one salvage therapy study, response rates to RTV/IDV were significantly higher among patients with an IQ>2 for IDV as compared to those with an IQ<2 (25). In the Viradapt study, optimal PI concentrations were an important determinant of outcome regardless of whether treatment was guided by the results of genotypic testing or not (see above) (19). Thus, TDM could play an important role in optimizing PI dosing in salvage therapy.

Despite these encouraging results, several important caveats must be kept in mind. First, drug levels need to be drawn at the appropriate time, preferably just prior to the next dose. For optimum interpretation, the time at which the last dose was taken needs to be recorded accurately. Second, the IQ should be based on IC50 values adjusted for drug binding to plasma proteins (see above), but there is considerable controversy over how such adjustments should be made. Third, there are limited data regarding the extent to which drug levels can be boosted safely. For certain highly resistant isolates, achieving an IQ >2 might require boosting a particular PI to levels that would not be safe, or for which safety data do not exist. Lastly, a randomized trial of TDM for management of salvage therapy, the Pharmadapt study, failed to show a benefit of TDM over standard of care (81). However, several aspects of the study design were flawed (for example, drug levels were related to the IC50 of control viruses, not the patient's own virus). Pharmadapt should be considered as a first attempt to study the role of TDM, not as the final word on this important topic.

In conclusion, although TDM offers potential promise, it is likely to be helpful only for a subset of drugs used in the treatment of HIV infection. A number of logistical hurdles need to be overcome and more data need to be gathered before TDM can be considered a part of the standard of care. As with virus load testing and drug resistance testing, moving TDM into the clinic will require carefully done retrospective and prospective studies to demonstrate the clinical utility of this potentially important too

Table 1. Major Drug Resistance Mutations

Nucleoside RT Inhibitors

Non-nucleoside RT Inhibitors

Protease Inhibitors

Mutations shown are the major mutations most commonly selected by each drug. Additional minor mutations may also accumulate and confer higher levels of resistance to the drug. Likewise, mutations selected by one drug may confer cross-resistance to another drug, but patterns of cross-resistance are not indicated in this table. For a more complete list of mutations see Hirsch et al JAMA 2000;283:2417-26; the resistane database at www.hiv.lanl.gov, or the Stanford PR and RT sequence database http://hivdb.stanford.edu/hiv/.

Reference List

1. Palella FJ, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV outpatient study investigators. N Engl J Med. 1998;338:853-60.
2. Havlir DV, Hellmann NS, Petropoulos CJ, et al. Drug susceptibility in HIV infection after viral rebound in patients receiving indinavir-containing regimens. JAMA. 2000;283:229-34.
3. Stuyver L, Wyseur A., Rombout A, et al. Line probe assay for rapid detection of drug-selected mutations in the human immunodeficiency virus type 1 reverse transcriptase gene. Antimicrob Agents Chemother. 1997;41:284-91.
4. Eastman PS, Boyer E, Mole L, Kolber J, Urdea M, Holodniy M. Nonisotopic hybridization assay for determination of relative amounts of genotypic human immunodeficiency virus type 1 zidovudine resistance. J Clin Microbiol. 1995;33:2777-80.
5. Japour AJ, Mayers DL, Johnson VA, et al. A standardized peripheral blood mononuclear cell culture assay for the determination of drug susceptibilities of clinical human immunodeficiency virus-1 (HIV-1) isolates. Antimicrob Agents Chemother. 1993;37:1095-101.
6. Hertogs K, de Bethune MP, Miller V. A rapid method for simultaneous detection of phenotypic resistance to inhibitors of protease and reverse transcriptase in recombinant human immunodeficiency virus type 1 isolates from patients treated with antiretroviral drugs. Antimicrob Agents Chemother. 1998;42:269-76.
7. Petropoulos CJ, Parkin N, Limoli K, et al. A novel phenotypic drug susceptibility assay for human immunodeficiency virus type 1. Antimicrob Agents Chemother. 2000;44:920-8.
8. Tisdale M, Kemp SD, Parry NR, Larder BA. Rapid in vitro selection of human immunodeficiency virus type 1 resistant to 3'-thiacytidine inhibitors due to a mutation in the YMDD region of reverse transcriptase. Proc Natl Acad Sci.U.S.A. 1993;90:5653-6.
9. Schuurman R, Brambilla D, de Groot T, Boucher C. Second worldwide evaluation of HIV-1 drug resistance genotyping quality using the ENVA 2 panel [abstract 58]. Antiviral Therapy. 1999;4 (suppl 1 ).
10. Qari SH, Respess R, Weinstock H, et al. A comparative analysis of Virco Antivirogram and ViroLogic PhenoSense phenotypic assays for drug susceptibility of HIV-1 [abstract 62]. Antiviral Therapy. 2000;5 (suppl 3).
11. Harrigan PR, Hertogs K, Larder B. Worldwide variation in antiretroviral phenotypic susceptibility in untreated individuals [abstract 455]. 8th Conference on Retroviruses and Opportunistic Infections, February 4-8, 2001, Chicago, IL. 2001.
12. Lanier ER, Hellmann N, Scott J, et al. Determination of a clinically relevant phenotypic resistance "cutoff" for abacavir using the phenosense assay [abstract 254]. 8th Conference on Retroviruses and Opportunistic Infections, February 4-8, 2001, Chicago, IL. 2001.
13. Kempf D, Brun S, Rode R, et al. Identification of clinically relevant phenotypic and genotypic break-points for ABT-378/r in multiple PI-experienced, NNRTI-naive patients [abstract 89]. Antiviral Therapy. 2000;5 (suppl 3):70-1.
14. Kuritzkes D, Sevin A, Young B, et al. Effect of zidovudine resistance mutations on virologic response to treatment with zidovudine/lamivudine/ritonavir: genotypic analysis of HIV-1 isolates from ACTG 315. J Infect Dis. 2000;181:491-7.
15. Larder BA, Kemp SD, Hertogs K. Quantitative prediction of HIV-1 phenotypic drug resistance from genotypes: the virtual phenotype (VirtualPhenotype) [abstract 63] . Antiviral Therapy. 2000;5 (suppl 3):49.
16. Graham N, Peeters F, Verbiest W, Harrigan R, Larder B. The virtual phenotype is an independent predictor of clinical response [abstract 524]. 8th Conference on Retroviruses and Opportunistic Infections, February 4-8, 2001, Chicago, IL. 2001.
17. DeGruttola V, Dix L, D'Aquila R, et al. The relation between baseline HIV drug resistance and response to antiretroviral therapy: re-analysis of retrospective and prospective studies using a standardized data analysis plan. Antiviral Therapy. 2000;5:41-8.
18. Durant J, Clevenbergh P, Halfon P, et al. Drug-resistance genotyping in HIV-1 therapy: the VIRADAPT randomized controlled trial. Lancet. 1999;353:2195-99.
19. Durant J, Clevenbergh P, Garraffo R, et al. Importance of protease inhibitor plasma level in HIV-infected patients treated with genotypic-guided therapy: pharmacological data from the Viradapt Study. AIDS. 2000;14:9.
20. Baxter JD, Mayers DL, Wentworth DN, et al. A randomized study of antiretroviral management based on plasma genotypic antiretroviral resistance testing in patients failing therapy. AIDS. 2000;14:F83-93.
21. Tural C, Ruiz L, Holtzer C, et al. Utility of HIV genotyping and clinical expert advice-The Havana Trial [abstract 434]. 8th Conference on Retroviruses and Opportunistic Infections, February 4-8, 2001, Chicago, IL. 2001.
22. Cohen C, Hunt S, Sension M, et al. Phenotypic resistance testing significantly improves response to therapy (Tx): a randomized trial (VIRA 3001) [abstract 237]. 7th Conference on Retroviruses and Opportunistic Infections, January 30-February 2, 2000, San Francisco, California. 2000;[Abstract]
23. Meynard JL, Vray M, Morand-Joubert L, et al. Impact of treatment guided by phenotypic or genotypic resistance tests on the response to antiretroviral therapy: a randomized trial (NARVAL, ANRS 088) [abstract 85]. Antiviral Therapy. 2000;5 (suppl 3):67
24. Brun-Vézinet F, Race E, Descamps D, et al. Differences between genotype and phenotype in the NARVAL trail, ANRS 088 [abstract 100]. Antiviral Therapy. 2000;5 (suppl 3):78-9.
25. Kempf D, Hsu A, Jiang P, et al. Response to ritonavir (RTV) intensification in indinavir (IDV) recipients is highly correlated with virtual inhibitory quotient [abstract 523]. 8th Conference on Retroviruses and Opportunistic Infections, February 4-8, 2001, Chicago, IL. 2001.
26. Hirsch MS, Brun-Vézinet F, D'Aquila RT, et al. Antiretroviral drug resistance testing in adult HIV-1 infection: recommendations of an International AIDS Society-USA Panel. JAMA. 2000;283:2417-26.
27. Little SJ, Routy JP, Daar ES, et al. Antiretroviral drug susceptibility and response to initial therapy among recently HIV-infected subjects in North America [abstract 756]. 8th Conference on Retroviruses and Opportunistic Infections, February 4-8, 2001, Chicago, IL. 2001.
28. Busch MP, Lee LL, Satten GA, et al. Time course of detection of viral and serologic markers preceding human immunodeficiency virus type 1 seroconversion:implications for screening of blood and tissue donors. Transfusion. 1995;35:91-7.
29. Ling AE, Robbins KE, Brown TM, et al. Failure of routine HIV-1 tests in a case involving transmission with preseroconversion blood components during the infectious window period. JAMA. 2000;284:210-4.
30. Kassler WJ, Haley C, Jones WK, Gerber AR, Kennedy EJ, George JR. Performance of a rapid, on-site human immunodeficiency virus antibody assay in a public health setting. J Clin Microbiol. 1995;33:2899-902.
31. Finzi D, Hermankova M, Pierson T, et al. Identification of a reservoir for HIV-1 in patients on highly active antiretroviral therapy. Science. 1997;278:1295-300.
32. Wong JK, Hezareh M, Gunthard HF, et al. Recovery of replication-competent HIV despite prolonged suppression of plasma viremia. Science. 1997;278:1291-5.
33. Chun TW, Stuyver L, Mizell SB, et al. Presence of an inducible HIV-1 latent reservoir during highly active antiretroviral therapy. Proc Natl Acad Sci.U.S.A. 1997;94:13193-7.
34. Allain JP, Laurian Y, Paul DA, et al. Long-term evaluation of HIV antigen and antibodies to p24 and gp41 in patients with hemophilia. Potential clinical importance. N Engl J Med. 1987;317:1114-21.
35. Phillips AN, Lee CA, Elford H, et al. p24 antigenemia, CD4 lymphocyte counts and the development of AIDS. AIDS. 1991;5:1217-22.
36. Fahey JL, Taylor JMG, Detels R, et al. The prognostic value of cellular and serologic markers in infection with human immunodeficiency virus type 1. N Engl J Med. 1990;322:166-72.
37. Jackson JB, Drew J, Lin HJ, et al. Establishment of a quality assurance program for human immunodeficiency virus type 1 DNA polymerase chain reaction assays by the AIDS Clinical Trials Group. J Clin Microbiol. 1993;31:3123-8.
38. Zazzi N, Romano L, Catucci M, et al. Low human immunodeficiency virus type 1 (HIV-1) DNA burden as a major cause for failure to detect HIV-1 DNA in clinical specimens by PCR. J Clin Microbiol. 1995;33:205-8.
39. Sauviago S, Barlet V, Guettari N, et al. Standardized nested polymerase chain reaction-based assay for detection of human immunodeficiency virus type 1 DNA in whole blood lystates. J Clin Microbiol. 1993;31:1066-74.
40. Mallet F, Hebrard C, Brand D, et al. Enzyme-linked oligosorbent assays for detection of polymerase chain reaction-amplified human immunodeficiency virus type 1. J Clin Microbiol. 1993;31:144-1449.
41. Whetsell AJ, Drew JB, Milman G, et al. Comparison of three nonradioisotopic poylmerase chain reaction-based methods for detection of human immunodeficiency virus type 1. J Clin Microbiol. 1993;30:845-53.
42. Elbeik T, Charlebois ED, Nassos P, et al. Quantitative and cost comparison of ultrasensitive human immunodeficiency virus type 1 RNA viral load assays: bayer bDNA quantiplex versions 3.0 and 2.0 and Roche PCR amplicor monitor version 1.5. J Clin Microbiol. 2000;38:1113-20.
43. Lin HJ, Pedneault L, Hollinger FB. Intra-assay performance characteristics of five assays for quantification of human immunodeficiency virus type 1 RNA in plasma. J Clin Microbiol. 1998;36:835-9.
44. Erice A, Brambilla D, Bremer J, et al. Performance characteristics of the QUANTIPLEX HIV-1 RNA 3.0 assay for detection and quantitation of human immunodeficiency virus type 1 RNA in plasma. J Clin Microbiol. 2000;38:2837-45.
45. Bartlett J, DeMasi R, Dawson D, Hill A. Variability in repeated consecutive measurements of plasma HIV RNA in persons receiving stable nucleoside reverse transcriptase inhibitor therapy or no treatment. J Infect Dis. 1998;178:1803-5.
46. Deeks SG, Coleman RL, White R, et al. Variance of plasma human immunodeficiency virus type 1 RNA levels measured by branched DNA within and between days. J Infect Dis. 1997;176:514-7.
47. Yen-Lieberman B, Brambilla D, Jackson B, et al. Evaluation of a quality assurance program for quantitation of human immunodeficiency virus type 1 RNA in plasma by the AIDS Clinical Trials Group Virology Laboratories. J Clin Microbiol. 1996;34:2695-701.
48. Nolte FS, Boysza J, Thurmo