Andrew Zolopa, MD, Stanford University
- transmission of drug resistance is a growing concern
- how long does drug resistance persist after transmission?
- the effect of the patient (host immune and genetic) makeup on the development of resistance: a new area in HIV
- resistance is still not fully understood; resistance mutations accumulate when on HAART and viral load is detectable
- nucleoside resistance: the effect of 3TC resistance on resistance to other nukes
- Tenofovir (Viread) resistance patterns and predictors of response
- NNRTI hyper-susceptibility
- use of ritonavir to boost other protease inhibitors: tipranavir resistance profile, predicting Kaletra resistance, indinavir+ritonavir dosing and phenotypic resistance cutoffs
- discordant resistance test results: you may not get the same results from submitting a sample for resistance tests
- "virtual phenotype" testing
- what is "virtual IQ" ?
- can Kaletra be taken once per day ?
At this year's conference there were a wide range of interesting abstracts dealing with HIV drug resistance. Although there were no "block buster" studies, some of the highlights included intriguing studies that demonstrate that host factors can influence the evolution of drug resistance. New studies that support the clinical utility of the "inhibitory quotient" or IQ as a prognostic tool. Additional studies on the impact of NNRTI hypersusceptibility on response to NNRTI-containing regimens as well as new insights on the evolution of drug resistance in newly infected individuals. In the following review we will focus on these highlights and try to place them into a practical clinical context for the reader.
Transmission of drug resistance: A growing problem but not everywhere
Investigators from the CDC presented data on drug resistance in a cohort of recently diagnosed and untreated HIV infected individuals as part of a 10 US city survey. They report on 1078 individuals and showed that there is an over all increase in the prevalence of resistance over the years of the study 1998-2000. Data from the poster presentation revealed that resistance as measured by genotype increased from 5.5% in 1998 to 10.7% for any antiretroviral with increases in resistance for all drug classes. Of note, in a subgroup who appeared to have been recently infected as they entered the cohort (by "detuned" ELISA) the over all prevalence of resistance was not different compared to the rest of the cohort with more chronic infection. [abstract 372]
Bob Grant and colleagues reported on increasing rates of drug resistance in newly infected individuals in San Francisco. The overall prevalence in this selected group increased from 10.7% in 1996 to 27.6% in 2001. The increases were particularly striking for resistance to NNRTI's which increased from 0% to 17% while resistance to PI's over this time interval increased from 0% to 10%. The investigators modeled the relative efficiency of transmission of resistant viruses and suggested that the more rapid raise in NNRTI resistance compared to PI resistance may be in part due to the fact that NNRTI resistant viruses may be relatively fit and able to better compete with wild type in transmission events. Further studies will be required to support the model presented. [abstract 368]
The overall prevalence of drug resistance in treatment naive infected individuals reported by the CDC study is somewhat less then other studies of US cities. However, these subjects are chronically infected compared to the cohort reported by Grant and also the multi-center cohort reported by Susan Little and colleagues. Furthermore, the CDC cohort has a more diverse cohort in terms of HIV risk factors and other socio-demographic factors that also likely play a role in the differences in observed prevalence. The important point is that this is the second largest multi-city study that supports the fact that drug resistance is increasing in treatment naive individuals in this country. Further evidence supporting the role of resistance testing in naive patients who are considering starting antiretroviral therapy.
In contrast to these studies from the US, studies from other countries including France, Australia and Spain do not demonstrate increasing transmission of drug resistance. The factors that explain these differences are not clear but part of the differences may have to do with immigration patterns of HIV infected individuals from the developing world. For example in the French report, 15% of the most recent new infections involved non-clade B HIV-1, presumably occurring within immigrant populations. It is likely that these populations have had less access to antiretroviral therapy and therefore less resistance is seen. These reports taken together illustrate both the heterogeneity and the overlapping nature of the global HIV epidemic. [abstracts 369,370,371]
How long does transmitted virus persist?
A commonly asked question by clinicians is how long does a resistant clone persist in an untreated individual who is infected with a resistant virus? Susan Little presented an interesting longitudinal analysis of 6 untreated individuals from the cohort of newly infected individuals she as previously reported. In these individuals she demonstrated that NNRTI resistance persisted well over a year after infection in 5/6 subjects. These findings support earlier case reports of the persistence of the K103N mutation. Furthermore, the results add support to the model of transmission of resistant virus in San Francisco by Grant et al. [see above]. It appears that NNRTI resistant viruses that are transmitted can compete relatively well against wildtype virus in newly infected subjects and are therefore able to persist despite the lack of selective pressure of antiretroviral treatment.
Dr. Little concluded her presentation with the recommendation that resistance testing be considered in newly infected individuals even up to a year after their presumed date of infection. It is interesting to note that the French government has already decided to pay for a baseline resistance test in patients who are felt to have been infected within a year prior to consideration for antiretroviral therapy. [personal communication Dr. Vincent Calvez, MD]
Immune surveillance and Host factors in the evolution of HIV drug resistance: A new paradigm for patient management? The future of individualized treatment
Most studies of drug resistance have focused on the role and impact of antiretroviral therapy on the development and evolution of resistance. At this year's conference there were several presentations that demonstrated the importance of immune surveillance and host factors on the evolution of drug resistance. From these studies it appears that HIV medicine is entering into a new paradigm, a paradigm in which knowledge of host factors will play an increasingly important role in the management of HIV infected individuals. I will highlight a few presentations as they relate to drug resistance.
Eric Daar presented a fascinating case of a newly infected individual who remained off antiretroviral therapy but was followed closely by Dr. Daar. The patient appeared to have been infected with at least two HIV strains one a relatively fit multi-drug resistant clone which persisted as the dominate strain for the first 110 days after infection and a second wild-type virus which quickly over took the MDR clone after day 110. In vitro measurements of replication capacity (fitness) did not explain this shift, that is, the MDR strain was measured to have a higher replication capacity. HLA typing suggested that the second strain emerged because of a shift in CTL epitopes allowing for the strain to escape immune surveillance and become the dominant strain. The implications of these observations are still to be fully realized.
In perhaps one of the most interesting poster presentations of the conference Simon Mallal's group from Western Australia presented a multivariate analysis that demonstrated that an individual's HLA type influenced the pattern and frequency of drug resistance mutations that developed in patients on antiretroviral therapies. Given the isolated geography of Western Australia, the group was able to study a nearly complete sample of the HIV infected adults in care and who are on ARV. The multivariate analysis clearly showed that HLA A and B types were independent predictors of drug resistance mutations at 8 codons in the reverse transcriptase sequence and 7 positions in the protease sequence. The authors conclude that these HLA-specific mutations represent CTL escape mutations and that knowledge of a patient's HLA type may help individualize antiretroviral treatment strategies. Clearly much more work will be required before HLA typing becomes a standard part of clinical evaluation for HIV infected patients who are about to start therapy but these results are very intriguing none-the-less. 
Using a different methodology from Mallal's group, Simon Frost and the group from UCSD presented a model which supports the idea that evolution of drug resistance in the protease sequence can not be fully explained by drug exposure. The authors suggest that HIV-specific immune responses are responsible for the variation noted at some protease codons. 
Does this mean that we should be HLA typing HIV patients as part of their evaluation for ARV? These papers suggest that this information may provide important prognostic information but further study is clearly warranted. None-the-less, it appears to this author that this will be part of the future of HIV care - things are likely to get more complicated!
Evolution of Drug Resistance while on Antiretroviral Therapy
A commonly held view on the evolution of drug resistance while on antiretroviral therapy is that of a single dominant clone which becomes increasingly resistant as it accumulates mutations to escape selective pressure of the ARV drugs. In an elegant study of patients on a failing PI-regimens, Alan Hance and colleagues demonstrated that this simple view of a single clonal evolution is probably an over simplification. Using a very sensitive real time PCR technique that can measure minority variants down to 0.1%, Hance analyzed serial plasma samples on patients with the L90M mutation and found a wide range of variants with different patterns of drug resistance mutations. These observations suggest that the evolution of drug resistance selected for under drug pressure is more like an extensively branching tree rather than a single branch.
In practical terms, having resistance tests with higher degree of sensitivity to minority populations may provide additional prognostic information that could help in managing antiretroviral therapy. This has not been demonstrated but the results of the Hance study suggest that there may be utility in knowing the full diversity of resistance present in an individual. Currently the LIPA assay is the only commercially available resistance assay with higher sensitivity to minority variants than either the more commonly used phenotype tests or sequence based genotype tests. What role this assay should play in clinical management is still undefined.
Several studies described the evolution of resistance in cohorts of patients who remain on ARV with ongoing HIV plasma viremia. Cloakley and colleagues nicely demonstrated that even in a highly select group of patients with ongoing viremia but whose viral loads remain below 1000 copies that there is still substantial evolution of resistance.  Several other authors showed evolution of drug resistance was frequent and developed to high levels for patients remaining on failing regimens. [566, 575]
The Problem of Cross Resistance Among the Nucleoside Analogs
There has been an increasing appreciation of the problem of cross resistance among the nucleoside reverse transcriptase inhibitors. This is largely related to the presence of mutations that have been classically associated with AZT resistance but are known to occur with other nucleosides. These mutations (at codons 41, 67, 70, 210, 215 and 219) are also referred to as "TAM's" for thymidine analog mutations or "NAM's" for nucleoside associated mutations. At least some of these mutations have been shown to increase the rate of pyrophosphorolysis, which reverses the polymerase reaction by removing the blocking nucleotide analog from the primer-template complex. In a plenary session, John Mellors gave a beautiful over view on the subject of nucleoside RTI resistance and he admonished the audience to "assiduously avoid NAM's". However, the best strategies to do that have yet to be defined.
In a very illustrative poster presentation, Jeanette Whitcomb and colleagues from ViroLogic displayed how the NAM mutations effect the phenotype for each of the drugs in the nRTI class, including tenofovir. She divided the nucleosides into those in which the 184V mutation accentuates (increases) the NAM-associated resistance (ABC, DDI, and DDC) and those in which the 184V attenuates (reduces) the NAM resistance (AZT, D4T, TDF and ADV). The graphs displayed in this poster will likely become the standard way of describing nucleoside cross resistance (John Mellors used them in his plenary lecture).
Tenofovir (TDF) Genotypic and Phenotypic predictors of response
Micheal Miller from Gilead Sciences delineated the effect of various genotypes and phenotypes on response to TDF in two Gilead sponsored trials (902, 907) in which TDF was added on to stable background therapy. The signature mutation for TDF is the 65R mutation which fortunately occurs rarely in treated patients (<5% of nucleoside experienced patients). In addition, the 69 insert multinucleoside resistant mutation appears to knock TDF out but the drug remains active against the other MNR pattern, the Q151M complex. Dr. Miller demonstrated that TDF maintained full activity in the presence of 1, 2 or 3 or more NAM mutations and that activity was only diminished if there were 3 or more NAM's inclusive of a 41L and/or 210W mutation. In general, the common triad of NAM mutations that seem to diminish but not completely nullify response to TDF were 41L, 210W, 215Y.
Using a recursive partitioning analysis, Miller and colleagues seemed to define the phenotypic "cutoff" for TDF. It appeared from this analysis that two cutoffs could be identified, one at 1.4 and one at 3.8. These will likely be used as defining complete response from a partial response (1.4) and partial response from no or limited response (around 4.0). These data should help practitioners use TDF in a more informed fashion.
Another study shows NNRTI hypersusceptibility improves virologic response
John Mellors on behalf of ACTG 398 study team presented an analysis of NNRTI hypersusceptibility (HS) and the impact on virologic response. HS was defined as a phenotype for EFV that was <0.4 fold-change compared to control. Recall that ACTG 398 enrolled PI-experienced patients (some of whom were also NNRTI experienced) into a randomized study of dual PI, versus single PI salvage. All patients also received EFV as part of the regimen. In the subgroup analyzed for this study, 29% (41/139) demonstrated HS to EFV at baseline. HS was shown to improved virologic outcome and was an independent predictor of virologic response at 48 weeks.
There are accumulating studies which support the idea that NNRTI hypersusceptibility can improve response to a NNRTI-containing regimens for the treatment of experienced patients. However, how this information can be used strategically in the management of ARV remains undefined. Although it is becoming increasingly clear that this in vitro characteristic measured by the ViroLogic phenotype has real implications in vivo, no one has advocated that NNRTI be used in the setting of treatment experience only to take advantage of this effect. Whether to use an NNRTI or not should be primarily based on patient-preference, side effect profile and dosing schedule considerations and not on the possibility of taking advantage of hypersusceptibility.
Hypersusceptibility is not isolated to NNRTI's it has also been shown for amprenavir. Viral isolates with the N88S mutation have previously been reported to show hypersusceptibility to amprenavir. Dr. Race and colleagues using the phenoscript phenotype assay showed that clinical isolates taken from the NARVAL trial with the D30N mutation associated with nelfinavir resistance along with the 88S/T seemed to be associated with hypersusceptibility to amprenavir and to a lesser extent ritonavir. The D30N mutation appeared to be associated with hypersusceptibility to ritonavir. Like NNRTI hypersusceptibility the clinical implications of these findings are not yet clear but could have implications for sequencing protease inhibitors. 
Boosted PI's and overcoming PI resistance
Tipranavir/r: A new boosted PI?
For some time now there has been considerable interest in tipranavir although precious little clinical data on the virologic activity of this novel PI in PI-experienced patients has been available. Doug Mayers and colleagues from Boehringer Ingelheim presented data from a phase 2 open label trial of NNRTI naive patients who were on a second failing PI regimen. Virologic response to tipranvir boosted with ritonavir using two doses of TPV/r were studied; 500/100mg BID and 1000/100mg BID. As expected this group of 41 patients had high level resistance to currently available PI's at baseline, measured both phenotypically and genotypically. Despite this level of resistance, all but 1 (2%) of baseline isolates were still susceptible to TPV (<4 fold-change) suggesting less cross resistance between TPV and other PI's. More importantly virologic response over 48 weeks was quite impressive with viral load declines of 2.5 log in both the low and high dose groups. Even in the subgroup who had major PI-resistance associated mutations including mutations at codons 46, 82, 84 and 90 viral load response was still impressive. Although not explicitly mentioned, it is assumed that these patients also received a NNRTI along with the TPV/r which surely contributes to response but much like the lopinavir/r studies in PI-experienced patients the results remain impressive for highly experienced patients such as those participating in this trial.
During the course of the study, 4 (10%) of patients developed resistance to TPV ( >10 fold-change in IC50), it appeared that resistance was associated with the development of the V82T mutation in association with mutations at codon 33. This is a relatively unique pattern of specific mutations although mutations at 33 are known to also be associated with resistance to LPV/r. The results of this small phase 2 trial are encouraging but we must await further study before it will be clear what role TPV/r will play in the treatment of PI-experienced patients.
Increasing Complexity of genotypic predictors of response to LPV/r (Kaletra)
Simple mutation scores were derived for LPV/r from relatively small studies of PI-experienced patients. This score was a simple summation of the number of mutations present from a list of 11 mutations (at codons 10,20,24,46,53,54,63,71,82,84,90) and was shown to have prognostic value in predicting response to LPV/r-containing regimens. As if this list of mutations wasn't already long enough we are now learning that there are additional PI-mutations which impact response to LPV/r.
Investigators from Abbott labs in collaboration with the French ANRS Resistance Group carefully analyzed the genotypic predictors of response to LPV/r as prescribed in the expanded access program in France. This data base contains nearly 800 patients who received LPV/r-containing regimens and who had a baseline genotype. Using logistic regression analysis the investigators identified additional protease mutations that were associated with response to LPV/r. They found that mutations at codons at 33, 36, 47 and 48 contributed to a diminished response to LPV/r.
In addition, the 50V mutation which is a signature mutation to amprenavir has been shown in vitro to be associated with significant resistance to LPV/r. In this study the 50V mutation was present in only 5 patients at baseline (<1%), however, none of these patients responded to LPV/r. It should be noted that other protease mutations were present along with the 50V in these patients.
Neil Parkin and colleagues showed that additional protease mutations (including those listed above) were significantly associated with LPV/r phenotype. One striking observation is that 25/26 isolates with 50V mutation had a LPV/r fold-change >10. Again supporting the potential of cross resistance between amprenavir and LPV/r. 
These studies illustrate the fact that with boosted PI's the number and patterns of mutations associated with resistance becomes very complex, making interpretation of the genotype a challenge even for the most experienced providers.
(editorial note: Dr Zolopa's last comment raises the concern that interpreting resistance testing for the regular HIV treater is perhaps too complex. The interpretations of resistance test results by some labs, which doctors often rely on to make decisions, may also not be so accurate. HIV is becoming an increasing complex disease which only highly qualified expertise can make reliable decisions in crucial & complex situations. Interpretation of resistance testing is I believe one of these situations where consultation with one of the key experts in this area is required, and how many doctors and patients have such access now. How many doctors are able to take the time to consult with such an expert?).
Phenotypic breakpoints for Indinavir+Ritonavir
Rice and colleagues reported on an observational cohort of patients followed at Stanford and Kaiser clinics who received IDV/RTV after prior PI therapy. They correlated response over 24 weeks to baseline IDV phenotype. It appeared that response to IDV/RTV was best in those patients with IDV fold-change <2.5 at baseline while patients with IDV fold-change >2.5 but <25 seemed to have some response and those with IDV fold-change >25 had little to no response to the boosted regimen. In addition, the investigators showed that the dose of IDV was an independent predictor of response. Patients who received 800mg IDV with either 100 or 200 mg of ritonavir had a better response than those treated with 400/400mg dosage. This supports predictions based on PK models. 
Phenotypic/genotypic discordance: What does it all mean?
The group at ViroLogic presented a careful analysis of phenotype/genotype discordance within the PhenoSense GT reports. Discordance was defined by one test indicating that a drug was sensitive while the other test indicated resistance. Discordant results were very common, over 70% of reports showing a discordant result for at least one drug. Discordant results were most commonly seen for DDI, DDC, ABC, 3TC and TDF (Tenofovir). Much of the discordant results could be attributed to negative interactions between mutations which result in a "sensitive" phenotype while the rules of genotype would classify as resistant. It is important to point out that some of the discordance could simply be a reflection of the genotype rules employed. Careful evaluation of these rules is required before we can make solid conclusions about the relevance and prevalence of discordance. An interesting comparison would be to use one of the standard genotype algorithms available and compare to the phenotype interpretation provided by ViroLogic. 
Moreover, what is most important is how these various patterns of genotypes and phenotypes correlate to virologic response in treated patients. In a sense this is the final arbiter of "truth".
Virtual Phenotype compares favorably to the real thing
There were two randomized studies comparing virtual phenotype to actual phenotype (Virco) presented at the conference. Both studies of treatment experienced patients showed that virologic response measured either at 24 weeks or 48 weeks was equivalent for the two arms of the study. [586, 589]
Based on these studies it appears that virtual phenotype has been established as a standard method for interpreting genotypic resistance results in a manner which is user friendly and reliable.
Inhibitory Quotients (IQ's): Combining drug levels and resistance
There were several presentations at this year's conference that evaluated the prognostic value of inhibitory quotients or IQ's. Generally, the IQ is a ratio of plasma drug concentration (eg Cmin) to a measure of viral resistance (eg IC50 ) for a given drug. In all the presentations, the IQ was measured for the PI component of the regimen in patients with prior PI experience.
There are now several versions of an IQ. The standard IQ [drug Cmin/IC50] is the most direct measure but there is also a virtual IQ and a normalized IQ. The virtual IQ replaces the measured IC50 with an estimate of IC50 based on the virtual phenotype. This virtual phenotype estimate then has to be corrected by a protein binding factor which can be measured in different ways. The normalized IQ is another measure of IQ developed by investigators at Virco. The normalized IQ is the measured Cmin divided by virtual phenotype FC and this ratio is then divided by the ratio of the expected population Cmin for the drug of interest divided by the biologic phenotype cutoff value. Although somewhat less straight forward compared to an IQ, the normalized IQ has the advantage over the virtual IQ in that it doesn't require the protein binding correction factor which has been subject to considerable debate.
Dr. Courtney Fletcher presented results from a retrospective analysis of 32 subjects from ACTG 359 in which IQ for saquinavir was measured and shown to be predictive of virologic response through the first 8 weeks of therapy on a SQV-containing regimen (editorial note: but was not predictive after I think it was 24 weeks. Observers asked why it was predictive at 8 weeks but not further along in time). The investigators looked at both an IQ which used the Cmin and an IQ that utilized an AUC measure of SQV and both were found to have prognostic value although the AUC IQ appeared to be slightly better. 
The normalized IQ was shown by Richard Hoetelmans of TiboTec-Virco to be predictive of 48 week virologic response to a LPV/r-based salvage regimen in a cohort of 52 PI-experienced patients. In multivariate analysis, the NIQ for LPV and baseline viral load were the two independent predictors of week 48 response. Thus the NIQ but not LPV Cmin itself was a better predictor of response than a virtual phenotype for LPV. 
Dr. Phillips and colleagues evaluated the vIQ in a cohort of 24 PI-experienced patients who received a salvage regimen containing LPV/r and APV. Virtual phenotype was determined from baseline genotype and drug levels were measured while the patient was on LPV/r/APV regimen for at least 3 months. It appeared that vIQ for both LPV/r and APV were associated with response. 
Investigators from Abbott laboratories presented data from a small randomized controlled trial of LPV/r 800/200 qday (once daily dosing) versus standard LPV/r 400/100mg BID in combination with d4T and 3TC in naive subjects. The investigators showed that the qD regimen was tolerated and grade 3/4 toxicities were similar across the two arms at 48 weeks. The two arms were similar in terms of virologic response with 74% of the qD achieving <50 copies by week 48 compared to 79% for standard BID dosing using an ITT analysis. In terms of PK profile the qD LPV produced an AUC and Cmax that was similar to BID dosing but the Cmin was only about 44% of that achieved by BID dosing. There was more inter-subject variation in Cmin in the qD arm as well. IQ were also somewhat lower and more variable in the qDay arm but still average IQ >40 in this arm.
This small study seems to support use of qD LPV/r in treatment naive patients without PI resistance. Given the somewhat lower and more variable Cmins of Q day dosing it is probably not advisable to consider this for PI-experienced patients who may have some degree of reduced susceptibility to LPV/r.
(editorial note: twice daily dosing appears preferable for more reliable viral response. But once daily dosing appears useful, particularly for special situations such as direct observed therapy or when adherence may be a special consideration and once per day therapy may be preferable for certain reasons. For treating children a once per day regimen with or without direct observed therapy may improve adherence when the parents are not careful about adherence).