Reports from the

5th Resistance Workshop on HIV Drug Resistance & Treatment Strategies

June 4-8, 2001 Scottsdale, Arizona

Summary of the 5th International Resistance Workshop: sessions 4, 5, and 6
      Written for NATAP by Andrew Zolopa, MD, June 23, 2001, Stanford University

--Phenotypic cutoffs or breakpoints including d4T, Indinavir (IDV/RTV) ...>
--Switching Regimens Every 3 months To Prevent Resistance: SWATCH Study ...>
--Kaletra (LPR/r) Resistance ...>
--Update on NNRTI Hypersusceptibility ...>
--Interpreting Genotype/Phenotype ...>
--Epidemiology: transmission of drug resistance to newly infected ...>

Introduction
At this year’s workshop there were beautiful presentations of the basic science of mechanisms of HIV resistance. Hopefully insights gained from this work will ultimately lead to new drugs with improved resistance profiles. There were fewer abstracts at this year’s workshop that related directly to pressing clinical issues. For example, we saw only a few abstracts that attempted to define resistance cutoffs or breakpoints for the antiretrovirals in current practice. Two important cutoffs should be defined for each of the drugs we use in practice, the first would define the cutoff between full activity of the drug and partial activity. A second cutoff would define partial response from no response. Another topic of interest to clinicians and patients is the use of pharmacologically enhanced ("PK boosted") regimens to overcome partial resistance, in particular protease resistance. As resistance testing becomes standard of care in HIV care, it has become clear that resistance is often not an all or nothing phenomenon. There are degrees of resistance (or more accurately, "reduced susceptibility" to drugs) and some degree of resistance can be overcome by increasing the levels of the drug in the patient ("PK boosting"). An example of this is seen with the new protease inhibitor, Lopinavir which appears to maintain good activity even when there is some degree of resistance to LPV/r and other protease inhibitors. This is likely due to the high drug concentrations that are achieved by the co-formulation with ritonavir. However, for most protease inhibitors used in practice today how much resistance can be overcome by PK boosting with ritonavir has not been clearly defined.

Session 4: Drug resistance, drug exposure and treatment response.

Phenotypic cutoffs or breakpoints

Indinavir+Ritonavir-- In a group of 42 protease experienced patients followed at the Stanford Positive Care clinic, Dr. Rice and colleagues using the Virologic PhenoSense assay, evaluated the phenotypic "breakpoint" for boosted indinavir regimens. The investigators explored what level of indinavir resistance (as measured by a fold change for indinavir IC50) would define a point where the boosted indinavir regimen no longer provided any virologic activity. Surprisingly the investigators found that at least some degree of antiviral activity was observed in patients with indinavir phenotypes of up to 25 fold change (FC). In fact, there were a few patients who had a response even with IDV FC>25. As expected, the best responses where seen in patients with FC near 1 (ie near wildtype) or in patients who had received NNRTI’s as a new class in addition to IDV/RTV. But even after controlling for these factors, Rice found that a majority of patients achieved at least a 0.5 log decline in viral load within the first 24 weeks with IDV phenotypes up to 25 FC. In addition, Dr. Rice showed that patients receiving a dose of 800mg of indinavir BID with 100-200mg of ritonavir had a better virologic response than those who had received 400mg of both drugs twice daily. Although limited by the small numbers this study should help providers decide when it is no longer feasible to use IDV/RTV. From these results it appears that IDV in a boosted regimens would provide at least some activity up to IDV FC of 25. Secondly the study provides evidence supporting PK modeling studies suggesting that the 800mg/200mg dosing of IDV/RTV provides is more potent that 400/400mg dosing regimens (the program book reports in patients with IDV IC50 <25 fold change 83% (10/12) using >400 mg twice daily IDV had >0.5 log reduction compared to 65% (13/20) using a 400-mg dose twice daily (p=0.26).

(Editorial note: Defining viral load response as a nadir viral load reduction over 24 weeks of >0.5 log, 73% of patients with < 25 fold resistant to IDV at baseline had a virologic response, compared to 30% (3/10) with > 25 fold IDV resistance).

Dr. Shulman and colleagues from Stanford, evaluated the phenotypic breakpoint for D4T. There is a growing appreciation of cross-resistance within the nucleoside class which limits their sequential use in salvage regimens. Resistance to D4T and DDI has been difficult to measure with phenotype assays as very small changes in IC50 values may have major implications for these drugs. In this study the Virologic phenotype was performed on stored specimens in 21 patients from ACTG 302 (a study of AZT experienced patients who where treated with D4T monotherapy). For patients with D4T fold change of 1.7 or greater at baseline there was limited or no response to D4T monotherapy suggesting that the cutpoint for D4T should be around 1.7 on the PhenoSense assay. Although not reflective of current treatment practices this small study allows for a relatively clear assessment of how baseline D4T phenotype influences virologic response to that drug. However, one potential problem with a cutoff at this level is that this is very close to limits of what the assay can reliably detect.

Dr. Francois Clavel and co-investigators from the NARVAL trial also tried to determine phenotypic breakpoints for antiretroviral drugs used in the standard of care (SOC) arm of NARVAL trial (a randomized controlled trial of genotype vs. phenotype vs. SOC). They found a variety of patterns relating virologic response to baseline phenotype for the different drugs. It appeared that some drugs had no relationship between phenotype and response, some had a linear relationship and indinavir and amprenavir seemed to have a threshold that defined response from no response. However, these results were not controlled for the virologic activity of the other drugs used in the multi-drug regimens the patients received in this trial and should therefore be seen as preliminary.

Dr. Richard Haubrich presented the 48 week results of the CCTG 575 study, a randomized controlled trial of phenotyping (Virologic) versus standard of care in picking new regimens for patients on failing first or second PI-based regimens. Overall there was no difference in the percent of patients who achieved viral loads of <400 or <50 copies at 6 or 12 months, nor was there a difference in the change in viral load at these time points. Dr. Haubrich showed that more DDI and D4T was used in regimens for the phenotype group and more abacavir was used in the SOC group. It was these differences in the nucleoside backbones employed in the two study groups, that the investigators feel explain the negative results of the study. In other words, the phenotype underestimated resistance to D4T and DDI and those drugs were then over used in the salvage regimens in that arm of the study.

Interestingly, in a subgroup of the more heavily pretreated patients, the phenotype group did have a better virologic response compared to the SOC. It should also be pointed out that the SOC group did very well in this study with 48% achieving <400 copies at 6 months. So the "negative" result of this study is not that the phenotype group did so poorly but rather it was that the SOC did so surprisingly well. In fact this rate of response in the SOC arm is superior to the response rates seen in some of the other resistance testing trials in the groups receiving the resistance test. None the less, this study points out the need to define phenotypes sensitivity cut offs by virologic response, further analysis of this study should help in this effort. For example, as has been mentioned previously phenotypic resistance to D4T is hard to detect. With this in mind Dr. Haubrich and colleagues used the AZT phenotype as a surrogate for D4T resistance and although the range of IC50’s for the two drugs are quite different there was a strong correlation between AZT phenotype and that of D4T (r2 =0.70). This observation has lead some to suggest that AZT phenotype be used as the resistance surrogate for the nucleoside class. Further study is required before we can define the expected activity of all nucleosides based on the phenotype of AZT.

Preventing resistance: Proactive switching therapy, a new approach?

In a interesting study called SWATCH, Dr. Martinez-Picado and colleagues from Barcelona demonstrated that "proactive" switching of ARV may lead to less virologic failure. In an update from their presentation at this year’s Conference on Retroviruses and Opportunistic Infections (CROI), treatment naïve patients where randomly assigned to one of three groups, D4T/DDI/EFV, AZT/3TC/NFV or switching every three months between the two regimens (switching group). In an interim intent-to-treat analysis, 70% of the switching group compared to 60% of the non-switching groups maintained viral loads of <400 copies at week 48 (p<0.05). There have been a number of studies that have evaluated the role of switching therapy for toxicity management but this study is the first to use ongoing switching (back and forth, in this case, between a NNRTI and PI-based regimens) in an attempt to improve virologic response. This study appears to confirm predictions from previous modeling experiments but we must await final analysis before drawing any further conclusions.

Refining our understanding of Resistance to Lopinavir/r (LPV/r)

There were several studies evaluating the resistance patterns associated with LPV/r. Investigators from Abbott laboratories have shown that >5 protease inhibitor mutations are usually required before the response to LPV/r –based ARV begins to wane. Dr. Kempf re-analyzed data from the Abbott trial of LPV/r in multiple PI-experienced subjects (study M98-957). He showed that mutations at codons 82, 54 and 10 plus a median of four additional mutations were associated with a lower response rate at 48 weeks of LPV/r therapy. Within the small group (n=20) with this pattern of mutations the LPV phenotype varied from 16-51 fold and the phenotype added additional information about likelihood of response. Interestingly, in this univariate analysis, major protease inhibitor resistance mutations at codons 84 and 90 were not predictive of 48 week response to LPV/r. In a much larger study presented by Dr. Calvez from the French ANRS HIV resistance group, the same pattern of mutations seemed to be most strongly related to response to LPV. In a clinical cohort of triple class experienced patients receiving LPV/r in a compassionate use protocol, Dr. Calvez and colleagues evaluated 700 patients. Baseline genotype and phenotype were obtained and response to LPV/r was monitored over 48 weeks. As has been shown in the previous Abbott studies, multiple PI-associated mutations appeared to be required before a significant decrease in response was seen. Virologic response was maintained in patients with up to 5 mutations and only diminished slightly in those with >6 mutations at baseline. Mutations at codons 10, 46, 54, and 82 were most strongly related to response as both univariate models and mutations at 54 and 10 were independent predictors in multivariate analysis.

In addition, the group from Abbott calculated a "virtual inhibitory quotient" or vIQ for LPV/r using the M98-957 study subjects and correlated this measure with response to LPV/r in the trial. The virtual IQ is calculated by measuring minimal plasma concentration of LPV (C min) and dividing by the virtual phenotype (calculated from the genotype by VIRCO) which is multiplied by the human serum corrected IC50. The formula for vIQ is: vIQ = Cmin/ vPT x IC50 (wt HIV, 50% human serum). A virtual phenotype (vPT) of between 10 and 15 appeared to give the best prediction of response, that is, 90% of patients with vPT below this level achieved a viral load of <400 copies at 48 weeks compared to 60% in the groups with vPT above this level. These results correspond very closely to the results from phenotype testing that have been previously reported by this group. Moreover, a vIQ of >10 gave the best discrimination between responders and non-responders. Again about 90% of patients with vIQ >10 had a viral load of <400 copies at 48 weeks compared to about 60% in those with vIQ’s of <10. However, it does not appear to this reviewer that the vIQ offered any advantage in predicting response over the vPT.

In contrast to the above studies that suggest multiple (ie >6) mutations are required for high level resistance to LPV/r, there were several studies presented at the meeting suggesting that the protease mutation I50V (classically associated with resistance to amprenavir) may play an important role as a resistance mutation to LPV/r. Neil Parkin of Virologic correlated the phenotypic resistance patterns of the currently available PI’s using a database with over 3000 clinical isolates. As expected there was a high degree of correlation between phenotypes for RTV and LPV (r2=0.93) and less correlation between SQV, AMP and NFV. Parkin also found that susceptibility to LPV was markedly reduced in viruses containing the I50V mutation even those with <6 other mutations previously correlated with response to LPV/r. These findings were supported by a study by Dr. Prado and colleagues from Barcelona and Boston. In in vitro serial passage studies with amprenavir, isolates with 50V and mutations at codons 10 and 46 along with gag cleavage mutations displayed high level resistance to LPV/r.

It appears likely that the I50V mutation is an important resistance mutation for LPV/r. The reason it was not determined to be associated with resistance in the Abbott trials is likely due to the fact that in those trials there were very few patients with amprenavir experience and therefore the prevalence of the I50V mutation was likely to be too low to show up as a predictor of response to LPV/r. Whether we will see 50V developing in patients on LPV/r –containing regimens is yet to be determined. If this should occur it would limit the likelihood that AMP-based regimens could serve as a salvage regimen for all LPV/r failures.

Update on NNRTI hypersusceptibility
      Contributions from Jules Levin

A brief summary is that it’s been shown in studies over the past 1.5 years that patients with more of certain NRTI mutations tend to be more sensitive to NNRTIs. And such patients see better viral load reductions and cd4 increases are seen over the first few months of treatment, although the viral load benefit fades over time. At this year’s Resistance Workshop, he results of two studies showed that the potential benefits of hypersensitivity to NNRTI may be overcome by adherence and a lack of adequate numbers of new drugs to which the patients are sensitive. This does not mean necessarily that hypersensitivity to NNRTIs cannot be useful under the right circumstances.

At last year’s conference there was much interest in the in vitro phenomenon of NNRTI hypersusceptibility. Hypersusceptibility is defined by the phenotype showing a fold change of <0.4 compared to the laboratory control virus (ie more sensitive to the drug then the control virus). In other words, hypersusceptibility is when a patient is more sensitive to a NNRTI (or any other HIV drug) than they are normally. It has been reported that the presence of NRTI mutations, following treatment with NRTIs, are the cause of hypersusceptibility to NNRTIs. At the 2000 Retrovirus Conference Deeks, Whitcomb, Hellman, and Petropoulos reported that NNRTi hypersusceptibility was seen in a significantly greater percentage of viruses from NRTI experienced patients (DLV 29%, EFV 26%, NVP 21%) compared to NNRTI naïve patients ((DLV 5%, EFV 9%, NVP 11%). They found that viruses with NNRTI hypersusceptibility had more NRTI mutations (n=4.5 vs 0.9). They found that phenotypic testing confirmed this but that fading NRTI resistance over time was associated with loss of NNRTI hypersusceptibility.

At last year’s conference, Dr. Haubrich reported similar findings when looking back at patients in study CCTG 575: hypersensitivity was associated with longer duration of NRTI use, prior AZT use, and AZT or abacavir resistance. He found reduced susceptibility to AZT and abacavir were associated with NNRTI hypersensitivity. But prior use of d4T, 3TC and ddI did not affect hypersensitivity. He showed that this hypersusceptibility was related to short term virologic control in patients treated with a NNRTI-based regimen. However, after a period of time the improved virologic response faded. Of the 73 patients who received EFV, the viral load reduction was 1.5 log±0.7 for patients who were hypersensitive to EFV compared to a 1.0±0.8 log reduction for patients not hypersensitive to EFV. The diference was statistically significant (p=0.03). a non-significant trend was seen at month 4: 1.5 log vs 1.0 (p=0.09). Month4 cd4 increase was greater for hypersensitive patients than non-hypersensitive patients. The month 2 viral load for patients receiving NVP (n=26) was similar, but not significant due to small numbers (1.7 log vs 0.9 log).

In a follow up study at this year’s Resistance Workshop, Dr. Kieser and colleagues showed that baseline hypersusceptibility did not appear to be associated with a better response to NNRTI-based therapy in a clinic-based cohort study. He looked at 14 patients non-hypersensitive to NNRTI EFV and 11 who were hypersensitive to EFV (change in EC50£0.4 fold for EFV or normal susceptibility). However, the non-hypersensitive patients received more active drugs (3 vs 1) than the patients who were hypersensitive. This suggests that the hypersensitive patients had a less potent regimen and could explain why patients with EFV hypersensitivity did not achieve as much viral suppression as the non-hypersensitive patients. Both groups had about a 1.5 log decrease in HIV viral load at 60 days, but the hypersensitive patients rebounded to baseline at 120 days while the non-hypersensitive group remained suppressed. The non-hypersensitive group failed therapy more rapidly than the hypersensitive group (254 days vs 74 days). Kieser concluded that the potential benefits of hypersensitivity may be outweighed by accumulation of resistance to other drugs. Hypersensitivity to EFV at baseline was associated with having the AZT mutations T215F/Y and M41L, and the T69D mutation. He suggests that the very cause of EFV hypersensitivity, more NRTI mutations, may prevent the patient from having adequate numbers of drugs to select for the next regimen, and this overrides the potential benefit to hypersensitivity. This suggests that when new drugs are available (PMPA, t-20) hypersensitivity to NNRTIs may be more useful.

In addition at this year’s Workshop, Dan Kuritzkes on behalf of the ACTG 370 study team showed that adherence to medication was more important than NNRTI hypersusceptibility in determining response in that trial. Recall that last year Vicki Johnson had shown that the presence of TAM’s (thymidine analogue mutations) was associated with better outcomes in the ACTG370 trial. These patients were on previous dual nucleoside regimens and put on a delavirdine containing ARV therapy. Dr. Kurizkes showed that the presence of TAM’s was a marker for better adherence on the dual nRTI’s regimens and this predicted better adherence to the DLV-containing regimen. NNRTI hypersusceptibility did not appear to be associated with response. In the 370 study only 13% of the patients were hypersensitive to delavridine, so if more patients had been hypersensitive it’s possible that it would have affected viral response. Further research is needed to explore the clinical usefulness of hypersensivity.

Session 5: Genotype/phenotype relationship: modeling and algorithms

In a new session this year, there were several presentations evaluating different methods and systems that are being developed to interpret genotypes. We now know from the HAVANA trial that expert advice is as important as having a genotype in improving outcomes to antiretoviral therapy in treatment experienced patients and that advice and genotype are additive in improving outcomes for patients. Although expert advice is important, there is no standardized method in making it available to the practitioner. In the ENVA 3 panel (see summary elsewhere), the participating labs were asked to rate drug activity based on the genotypes of the 5 isolates studied. Dr. Wensing from Utrect Medical center, demonstrated that there was wide variation in the opinions about drug activity. To illustrate this, she showed that for a wild-type protease sequence nearly 20% of the labs considered LPV/r not to be active while in the presence of a L90M, 30% of the labs considered NFV and SQV to still be active. This variation in interpretation was not specific to a particular drug, lab or technology employed. The investigators call for a standardized interpretation program analogous to having standardized kits to perform the genotyping.

Dr. Bob Shafer from Stanford compared the interpretation from three different rules-based algorithms, the RCG DAP, the French National Agency for AIDS research (ANRS) and the Stanford HIV database program. In comparing the interpretations from over 2000 sequences in the Stanford database, Dr. Shafer showed that over 80% of the time there were concordance between the three systems and 18.6% there were discordance. Nearly 3/4 of the discordance were related to interpretations involving DDC, ABC, DDI, D4T and amprenavir, that is, drugs for which there has been difficulty in determining resistance patterns (DDC.DDI, D4T) or newly available drugs (ABC, AMP). Dr. Shafer was optimistic that the various systems for interpretation would rapidly converge over the next year as the systems are refined.

Dr. Lanier from GlaxoSmithKline, used data from ABC intensification trials to compare the predictions from different commonly used interpretation programs in a meta-analysis. There was a wide range in the accuracy of the interpretation programs in predicting response to ABC in these trials. Both errors of predicting sensitivity but no response was seen ("false positive") and predicting resistance but response was seen ("false negative") were noted a quarter to one third of the time. Based on this analysis it appears that the interpretation programs for ABC require further refinement which supports the results of Dr. Shafer’s comparison.

There were presentations in this session evaluating other systems under development to interpret genotypes as well as interesting presentations of new statistical methods being applied to the complex genotypic datasets. This area of technology and statistical methodology is likely to evolve rapidly, however, for now much of this work is not directly applicable to clinical practice.

Session 6: Epidemiology and subtypes of drug resistance

Susan Little from UCSD updated the study she presented at this year’s CROI, evaluating drug resistance in newly infected patients from 9 North American cities. In the nearly 400 patients enrolled into the cohort thus far, the transmission of drug resistant HIV appears to be increasing. For patients enrolled from 1995 to 1998 the overall prevalence of resistance (measured by phenotype) was 3.8% and that rate has increased to 14% in the group enrolled in 1999-2000. The prevalence of drug resistance increased for all classes of drugs, up to 8% for the nucleoside inhibitors and NNRTI’s, up to 9% for PI’s and most alarming up to 6% of newly infected patients in the 1999-2000 group had multi-drug resistant virus (ie resistance to 2 or more classes). Dr. Little again showed that the response to therapy appeared to be diminished in the group with resistant virus, measured both as a decreased time to suppression and earlier time to failure.

The CDC presented a study of drug resistance patterns in over 700 drug naïve subjects enrolled from 10 US cities. In this study, drug naïve subjects who sought HIV testing between 1997 to 2000 were evaluated by HIV genotype for the presence of drug resistance mutations. To evaluate the group with newly acquired infection the investigators used the "detuned" ELISA antibody assay which identifies those subjects who were infected in the past 4 to 6 months prior to testing. Overall the investigators found nearly 8% of subjects had a least one major drug resistance mutation. The prevalence of drug resistance was higher in men who have sex with men compared to heterosexuals or injection drug users. There was no apparent difference in the prevalence of drug resistant mutations in the group with recent infection compared to the group with more chronic infection. This study nicely illustrates that transmission of HIV drug resistance is going to vary by risk group, ethnicity, location and time and this variability is another reflection of the fact that we do not have one monolithic HIV epidemic in this country but multiple overlapping but distinct epidemics.

Dr Hanna and colleagues from the Massachusetts General Hospital evaluated the response to antiretroviral therapy in a group of 88 consecutive drug naïve subjects presenting for treatment in 1999. The investigators found that 18% of these chronically infected treatment naïve subjects had at least one mutation associated with drug resistance. The response to therapy was slower in the group with drug resistant virus compared to those with wild-type virus but all subjects with drug resistant virus achieved viral loads of <50 copies by 40 weeks. These results seem to partially support the findings of Dr. Little but as these authors suggest further study of the impact of drug resistance transmission is needed.

There were several studies presented on the resistance patterns and response to therapy in cohorts with non-clade B HIV infection. It is important to keep in mind that most of the studies to date evaluating HIV resistance are in clade B virus yet worldwide clade B only represents about 3% of infected persons. Clearly much more study of resistance and response to therapy is needed in non-clade B virus particularly as drug therapy is expanding to populations infected with non-clade B virus. Dr. Loveday and colleagues from the Royal Free Hospital in London conducted a case control study evaluating response to therapy in non-clade B virus compared to clade B. The virologic and immunologic response to antiretroviral therapy was nearly equivalent in the two groups at 48 weeks. Other studies presented at the conference showed that the pattern of mutations and polymorphisms appears to vary somewhat between the different HIV clades. The impact of this variation on response to therapy and evolution of resistance requires further study.