HIV Articles  
Back 
 
 
Defining HIV Susceptibility to New Antiretroviral Agents-Darunavir EDITORIAL
 
 
  Journal of Infectious Diseases Sept 2007
 
Richard Haubrich
 
Department of Medicine, Division of Infectious Diseases, University California San Diego, San Diego, California
 
(See the brief report by Mitsuya et al.below)
 
Financial support: National Institutes of Health (grant K24-AI064086); UCSD AIDS Clinical Trials Unit (grant AI 27670); the UCSD Center for AIDS Research (grant SP30 AI36214); California Universitywide AIDS Research Program: California Collaborative Treatment Group (grant CH05-SD-607-005).
 
As the number of agents that are used to treat HIV increases, the decisions regarding which antiretroviral agents to select in each clinical setting become more complex. The choice of the components of the regimen for patients who have an extensive treatment history relies on a careful assessment of (1) that past treatment history, (2) the response to prior agents, and (3) the results of past and current resistance tests [1]. In the absence of transmitted resistance to a drug, the virologic response to a drug from a class that a patient has used is almost always less than the response to a drug from a class that a patient has never used [2, 3]. The purpose of resistance testing is to help us to define which agent(s) of a class will have the best residual activity [1, 4, 5]; however, the interpretation of the resistance-test results can be difficult. For example, in the protease inhibitor (PI) class, the patterns of the PI resistance-associated mutations are complex and vary from drug to drug. PI resistance-associated mutations can be identified by in vitro selection experiments, by cataloging the appearance of new mutations during therapy, or by relating virologic response to a drug in patients with existing mutations. Ideally, the genotypic algorithm that defines the activity of each drug is based on large data sets, in which the patterns of the mutations from a baseline genotype can be interpreted in relation to HIV-RNA responses to a new regimen containing the drug of interest [6]. The limitations of the earlier studies are that the sample sizes are small, compared with the large number of possible genotypic mutational patterns, and that most of their analyses define the relative, as opposed to the actual, potency of a drug in the regimen.
 
When new antiretroviral drugs are submitted to regulatory authorities for review, information about the effects that both genotypic and phenotypic resistance to drugs has on the predicted response to the drug is provided [7]. Indeed, the agents approved for use in the past several years have had, at the time of market release, significantly more drug-resistance information available than have earlier agents. However, the mutation patterns that are present in the study populations may not fully define all of the mutations that could lead to a reduced response to an agent. Additional data, from clinical trials or observational databases, can augment the interpretation algorithms for new agents. In this issue of the Journal, Mitsuya et al.'s brief report [8], which suggests that the mutational scores derived from the clinical trials to date may be incomplete for the interpretation of clinical resistance, represents a step in the right direction for the recently approved drug darunavir (DRV).
 
The preliminary analyses from the POWER studies found 11 mutations (V11I, V32I, L33F, I47V, I50V, I54L, I54M, G73S, L76V, I84V, and L89V) that were associated with a reduced response to a DRV-containing regimen [9]. These analyses used a plasma HIV-RNA level of <50 copies/mL at 24 weeks as the outcome variable. DRV resistance-associated mutations were defined as those which were present in subjects who were less likely to achieve viral suppression, compared with the overall response of the entire population; increasing the number of these mutations was associated with a reduction in phenotypic susceptibility and virologic response. However, the accumulation of PI mutations defined by the International AIDS Society-USA [10] was not as closely linked to viral responses.
 
These analyses have a number of limitations: (1) they do not account for the effects of other drugs in the regimen (DRV was initiated with an optimized background of nucleoside reverse-transcriptase inhibitors [NRTIs], with or without enfuvirtide); (2) they do not weight mutations that might have a bigger impact on virologic response; and (3) the selected end point (viral suppression at 24 weeks) may have been more dependent on the regimen containing adequate numbers of active drugs rather than on the intrinsic potency of DRV against the PI-resistant isolates. Further sensitivity analyses (using a genotype- or phenotype-sensitivity score [6, 11-13]) should be performed with the data set, to account for other agents in the regimen and to explore other end points. For these analyses, using intent-to-treat data, which might ascribe failure to administrative reasons in addition to HIV-RNA levels of >50 copies/mL, may be less informative than as-treated data. For example, in the package insert for DRV, a different analysis used as-treated data to relate the number of amino acid changes at positions 30, 32, 36, 46-48, 50, 53, 54, 73, 82, 84, 88, or 90 (any change was counted) to the 1-log10-copies/mL reduction of HIV RNA at week 24 [14]; for patients with 0-4 PI mutations, the response was 81%, for patients with 5-6 mutations, it was 76%, and for patients with 7 or more mutations, it was 21%. Thus, with the same data set but with different analytic methods, 2 different algorithms have been generated; potentially, for an individual patient, the 2 algorithms could give 2 different estimations of the likelihood of response to a regimen that contains DRV.
 
How do the results reported by Mitsuya et al. contribute to our understanding of susceptibility and response to DRV? The investigators conducted a retrospective cohort analysis using 2 databases. The prevalence of DRV resistance-associated mutations, based on the list from De Meyer et al. [9], was determined for both data sets. Not surprisingly, the prevalence of these resistance-associated mutations in PI-naive patients was low in both data sets, but the prevalence among PI-experienced patients was high, nearly 30% in 1 data set. As expected, the increase in exposure to PI and the use of amprenavir increased the number of DRV resistance-associated mutations. Mitsuya et al. comment that the majority of patients in these cohorts would have fewer than 3 DRV resistance-associated mutations and would likely have a good response to DRV-based therapy, which is certainly good news for the population of triple-class-experienced patients.
 
Importantly, the authors also found 17 additional mutations that were statistically associated with the presence of at least 1 of the DRV mutations in De Meyer et al.'s list, which implies that there may be other DRV resistance-associated mutations that, when present, may impact the response to DRV-based regimens. One explanation for the identification of different mutations is that the populations evaluated by Mitsuya et al. were different from the subjects evaluated in the POWER studies. Most notably, the proportion of subjects with exposure to 1 or 2 PIs was 61%-81% in the cohort studied by Mitsuya et al., compared with 95%-100% in the POWER studies [15, 16]. In addition, the populations studied by Mitsuya et al. used more nelfinavir (48%) and less amprenavir (8%). Likely, the degree of DRV resistance in the Mitsuya et al. cohorts may be less than that in other clinic populations, especially those in whom there is greater breadth and diversity of PI use.
 
One major limitation of the Mitsuya et al. analysis is the lack of clinical-response data that would help us to evaluate whether these additional mutations are associated with a reduced virologic response to DRV-containing regimens. Further work, using data sets with genotypic- and phenotypic-resistance assays and virologic-response data, is needed to validate the De Meyer et al. set of resistance-associated mutations and to see how the additional Mitsuya et al. mutations improve treatment-response predictions. Cohort databases are powerful research tools, but they do not allow the same rigor that prospective clinical trials do; they have neither the same level of data-quality assurance nor the breadth of clinical covariates, such as past treatment history, toxicity, and adherence. The DUET studies (large randomized studies of TMC125, a new NNRTI active against NNRTI-resistant virus) used DRV in all the regimens; the further refinement of genotypic algorithms could be an added benefit of secondary analyses of these data sets.
 
Because current algorithms have been designed for single drugs and have not been compared across available drug treatment options, De Meyer et al.'s and Mitsuya et al.'s mutation sets are, at present, incomplete, and this complicates their use in the clinic. Ideally, genotypic-resistance algorithms would report the relative susceptibility of DRV compared with the predicted susceptibility of other PIs. For example, it is not clear whether the presence of 3 mutations from the DRV set, rather than from another drug, such as tipranavir, might predict a better response if DRV were used in a regimen. In this case, phenotype assays may add additional information about the relative activity of the available PIs [1].
 
Even after the clinician identifies the optimal PI, many questions remain concerning the selection of other drugs for the regimen. Even if the resistance test suggests that DRV is the optimal PI, it is still unclear how much activity can be ascribed to that drug and how many additional agents might be needed to achieve viral suppression. Many recent studies have confirmed that the addition of a new class of agents-including a fusion inhibitor (enfuvirtide), a CCR5 antagonist (maraviroc), or an integrase inhibitor (raltegravir)-greatly increases the likelihood of achieving HIV-RNA levels of <50 copies/mL [17-21]. Current guidelines now recommend that viral suppression be the goal of therapy, even in extensively treatment-experienced patients, and that at least 2, optimally 3, active agents be used in a new regimen [2]. But when can we consider DRV or other agents from recycled classes to be fully active agents? How many agents, and from which classes, need to be combined in the regimen? These and other questions remain.
 
Although recent studies of treatment-experienced patients have demonstrated the benefit of new agents, additional research is needed to evaluate strategies of regimen selection in heavily treated populations. To that end, the AIDS Clinical Trials Group is developing A5241, a randomized study that will attempt to answer 2 questions: (1) Are NRTIs necessary when more than 2 active agents are present in a new regimen? (2) Can a phenotypic susceptibility score be used to "add up" the relative activity of recycled agents and to define the threshold level at which sufficient activity is present in a regimen, so as to not add all possible new or recycled drug classes? This and similar studies will help to define the way forward into a new era of antiretroviral therapy in which high rates of viral suppression can be achieved in HIV-infected patients at all stages of the treatment continuum.
 
Prevalence of Darunavir Resistance-Associated Mutations: Patterns of Occurrence and Association with Past Treatment
 
JID Sept 2007
 
Yumi Mitsuya,1 Tommy F. Liu,1 Soo-Yon Rhee,1 W. Jeffrey Fessel,2 and Robert W. Shafer1
 
1Division of Infectious Diseases, Department of Medicine, Stanford University, Stanford, and 2Kaiser-Permanente Medical Care Program-Northern California, Oakland, California
 
(See the editorial commentary by Haubrich)
 
Eleven protease mutations have been associated with reduced susceptibility to darunavir (DRV). We examined the prevalence and covariates of these mutations in 2 populations. Thirty percent of 1175 Northern California patients and 24% of 2744 non-California patients in the Stanford HIV Drug Resistance Database had viruses with 1 or more mutations associated with resistance to DRV. In multivariate analyses, the number of DRV resistance-associated mutations depended on the number of previous protease inhibitors (PIs) administered and on amprenavir/fosamprenavir treatment. Most PI-treated patients should respond favorably to DRV-based salvage therapy.
 
Darunavir (DRV, formerly TMC114) is a recently licensed protease inhibitor (PI) with in vitro activity against HIV-1 isolates resistant to other PIs and with clinical efficacy in the treatment of persons in whom multiple previous PI-containing regimens have failed [1, 2]. Much of the DRV-prescribing information is derived from the phase II POWER registration studies, in which different ritonavir-boosted doses of DRV, as well as a comparison ritonavir-boosted dose of PI, were used for salvage therapy [3].
 
In an adjunctive analysis of the POWER studies, De Meyer et al. identified, at 10 protease positions, 11 mutations associated with both a reduced in vitro susceptibility to DRV and a reduced in vivo virologic response to DRV salvage therapy [4]. Among subjects receiving DRV/r, the percentages of subjects with 0, 1 or 2, and 3 or more of the 11 DRV resistance-associated mutations who attained plasma HIV-1 RNA levels <50 copies/mL at week 24 were 60%, 45%, and <20%, respectively [4]. Although other covariates, such as baseline plasma HIV-1 RNA levels and the use of enfuvirtide, influenced virologic response, no PI resistance-associated mutations other than the 11 identified by De Meyer et al. were reported to influence the virologic response.
 
The prevalence of these 11 DRV resistance-associated mutations in PI-naive and PI-treated patients, the risk factors for the development of these mutations, and the other protease mutations with which these mutations occur have not been described. Therefore, we examined their prevalence and covariates in 2 mutually exclusive populations with different potential biases: patients in a large representative US clinic population and patients in published studies in the Stanford HIV Drug Resistance Database.
 
Patients, materials, and methods. HIV-1 protease sequences were examined in 2 mutually exclusive populations with a known history of PI treatment: (1) a population of PI-treated patients from 16 clinics of Kaiser-Permanente Medical Care Program-Northern California, from whom plasma samples were collected and submitted for genotypic resistance testing at Stanford University Hospital between 1998 and 2006, and (2) PI-treated patients described in published studies in the Stanford HIV Drug Resistance Database [5]. For viruses undergoing genotypic resistance testing at Stanford University, treatment histories were obtained from patients' charts and pharmacy records, as part of a collaboration approved by the institutional review board. For data from published studies, treatment histories were supplemented with requests for information from the studies' authors.
 
DRV resistance-associated mutations were defined by De Meyer et al. as the following 11 differences, at 10 protease positions, in the subtype B consensus protease sequence: V11I, V32I, L33F, I47V, I50V, I54L, I54M, G73S, L76V, I84V, and L89V [4]. Mutations that were present as part of an electrophoretic mixture (i.e., more than 1 peak was present at a position on the sequence electropherogram) were classified as mutations. Sequences containing a mixture of I54L and I54M, however, were considered to have only 1 DRV resistance-associated mutation at position 54. If more than 1 virus isolate with DRV resistance-associated mutations was taken from a patient, the isolate with the most DRV resistance-associated mutations was used to determine the frequency and patterns of DRV resistance-associated mutations.
 
Results.
 
The clinic population included 1847 patients, of whom 1175 had received 1 or more PIs. The database population included 11,697 patients, of whom 2744 had received 1 or more PIs. Each of the 11 DRV resistance-associated mutations was uncommon among sequences from PI-naive individuals, having rarely occurred at a prevalence of >0.5% in any of the 8 most common subtypes ([5], http://hivdb.stanford.edu/cgi-bin/MutPrevBySubtypeRx.cgi). Notable exceptions include V11I, which was present in 0.6%, 1.9%, 0.6%, and 2.3% of CRF01_AE, CRF02_AG, D, and G isolates, respectively, and L33F, which was present in 1.0% and 1.2% of A and AE isolates, respectively. We note that L33V and other substitutions at position 89 (particularly L89M) were highly polymorphic, with L89M being the consensus residue in multiple subtypes.
 
In contrast to their almost complete absence in isolates from PI-naive persons, the 11 DRV resistance-associated mutations did occur in PI-treated persons. Among the 1175 Northern California patients, 29.8% had 1 or more DRV resistance-associated mutations, including 25.7% who had 1 or 2 mutations and 4.1% who had 3-6 mutations. Among the 2744 patients in the Stanford HIV Drug Resistance Database, 23.7% had 1 or more DRV resistance-associated mutations, including 22.8% who had 1 or 2 mutations and 0.9% who had 3-6 mutations. Plasma HIV-1 RNA levels and CD4 counts were available for the 1175 Northern California patients. Among the 350 patients with DRV resistance-associated mutations, the median RNA level was 4.3 log copies/mL, and the median CD4 count was 220 cells/mm3. Among the 825 PI-treated patients without DRV resistance-associated mutations, the median RNA level was 4.0 log copies/mL, and the median CD4 count was 272 cells/mm3.
 
I84V, G73S, L33F, V32I, I54L, and I54M were the most common mutations in both populations (table 1); V11I, I47V, I50V, L76V, and L89V were the least common. The most common combinations of mutations included G73S+I84V, L33F+I84V, V32I+I47V, L33F+I54L/M, and I54L/M+I84V.
 

Prevalen-1.gif

Among the 1175 PI-treated clinic patients, 37%, 24%, 17%, and 22% received 1, 2, 3, and 4 or more PIs, respectively. Among the 2744 PI-treated database patients, 62%, 19%, 12%, and 8% received 1, 2, 3, and 4 or more PIs, respectively. The proportions of patients receiving each of the PIs were similar in both data sets. Among the pooled clinic and database patients, indinavir was used in 53%, nelfinavir in 48%, saquinavir in 34%, lopinavir in 9%, amprenavir and/or fosamprenavir in 8%, atazanavir in 2%, and tipranavir in 0.4%.
 
For the clinic population and the database population, individually, we created several multivariate regression models to characterize the relationship between past antiretroviral therapy and the number of DRV resistance-associated mutations. In both populations, the number of DRV resistance-associated mutations was independently positively associated both with the number of PIs previously received (P < 10-16) and with having previously received either amprenavir or fosamprenavir (P < 10-16). When we controlled for the number of PIs received, there was a negative association with having received nelfinavir (P < 10-4), in both populations, and with having received atazanavir (P < 10-5), in the clinic population.
 
Because PI resistance-associated mutations often occur in complex patterns, we performed a regression analysis to identify mutations that increased in prevalence with an increasing number of DRV resistance-associated mutations and that might therefore also be predictors of response to DRV therapy. For this analysis, we included only sequences containing at least 1 of the following nonpolymorphic PI resistance-associated mutations: D30N, V32I, M46I/L, I47A/V, G48V, I50L/V, I54A/V/L/M, L76V, V82A/F/L/S/T, I84V, N88D/S, and L90M, to exclude mutations that might appear to be associated with DRV resistance-associated mutations simply because they were more common in isolates with evidence of selective drug pressure.
 
A total of 131 mutations at 62 protease positions were present in viruses from 15 or more treated patients in the data set of pooled clinic and database patients. After using the method of Benjamini and Hochberg [6], to control the false-discovery rate at 0.05, we found that 17 mutations at 16 protease positions were significantly associated with the number of DRV resistance-associated mutations, including L10I (P < 2e-16), M46I (P < 2e-16), A71V (P < 2e-16), I72L (P < 2e-16), L90M (P < 2e-16), C67F (P = 3.5e-11), G16A (P = 3.93e-11), L63P (P = 8.5e-11), K55R (P = 6.8e-9), K43T (P = 7.8e-9), F53L (P = 9.8e-8), G73C (P = 1.4e-7), I62V (P = 8.0e-7), G73T (P = 3.9e-6), I85V (P = 2.2e-5), L10F (P = 2.0e-4), and Q18H (P = 4.0e-4). Figure 1 shows the results of the analysis for 9 of the most common non-DRV PI resistance-associated mutations, demonstrating the positive correlation with M46I and L90M but not with D30N, G48V, I54V, and V82A/F/I/T.
 
Figure 1. Correlation between number of darunavir (DRV) resistance-associated mutations and 9 additioinal PI resistance-associated mutations. The x-axis represents the no. of DRV resistance-associated mutations (with the n values, in parentheses, indicating the no. of patients), and the y-axis represents the frequency of the 9 PI resistance-associated mutations: D30N, M46I, I48V, I54V, V82A, V82F, V82I, V82T, and L90M. M46I and L90M were the only mutations that increased in frequency as the number of DRV resistance-associated mutations increased. M46I and L90M were significantly correlated with the number of DRV resistance-associated mutations, even after we controlled for multiple comparisons. None of the remaining mutations were positively correlated with the number of DRV resistance-associated mutations. The database and clinic population sequences were pooled for this analysis.
 
Discussion.
 
Because the DRV resistance-associated mutations are not among the most commonly occurring PI resistance-associated mutations, we sought to identify their prevalence, their patterns of co-occurrence, and their risk factors in a clinic population and in an online database of protease sequences from published studies. Our results show that 96% of PI-treated persons in a Northern California clinic population and 99% of PI-treated persons listed in the database have fewer than 3 DRV resistance-associated mutations and would therefore be expected to have a favorable response to DRV, with an 50% chance of achieving a plasma HIV-1 RNA level of <50 copies/mL by week 24 [4].
 
Not surprisingly, both the number of previously received PIs and previous amprenavir or fosamprenavir treatments were associated with an increased number of DRV resistance-associated mutations. Amprenavir and DRV are highly similar molecules, the only difference being that DRV has a fused bicyclic tetrahydrofuran [7]. Moreover, the drug resistance profiles are similar; 9 of the 11 DRV resistance-associated mutations also reduce amprenavir susceptibility [8].
 
Several lines of evidence suggest that the list of DRV resistance-associated mutations reported by De Meyer et al. may not be the only mutations influencing susceptibility to DRV and virologic response. First, attempts at selecting for resistance to DRV were unsuccessful in vitro when they began with the wild-type virus [1] but not when they began with viruses containing other PI resistance-associated mutations [9]. Second, mutations are often surrogates for other drug resistance-associated mutations. The rapid emergence of resistance to newly administered PIs has been reported to develop in patients having a pretherapy plasma virus that contains multiple PI resistance-associated mutations but that nonetheless is susceptible to the prescribed PI [9]. Finally, the clinical data supporting the current list of DRV resistance-associated mutations are based on a single-patient population. Common mutations such as M46I and L90M, which we found to be highly correlated with the number of DRV resistance-associated mutations, may have an effect on virologic outcome despite not having had a significant effect when added to a model that already contained 11 mutations with which they were highly correlated.
 
 
 
 
 
  icon paper stack View Older Articles   Back to Top   www.natap.org