Skip Navigation

Variable Prediction of Antiretroviral Treatment Outcome by Different Systems for Interpreting Genotypic Human Immunodeficiency Virus Type 1 Drug Resistance

  1. Andrea De Luca1,
  2. Antonella Cingolani1,
  3. Simona Di Giambenedetto1,
  4. Maria Paola Trotta2,
  5. Francesco Baldini2,
  6. Maria Gabriella Rizzo1,
  7. Ada Bertoli2,
  8. Giuseppina Liuzzi2,
  9. Pasquale Narciso2,
  10. Rita Murri1,
  11. Adriana Ammassari1,
  12. Carlo Federico Perno2 and
  13. Andrea Antinori2
  1. 1Institute of Clinical Infectious Diseases, Catholic University, and
  2. 2National Institute for Infectious Diseases “Lazzaro Spallanzani” Istituto di Ricovero e Cura a Carattere Scientifico, Rome, Italy
  1. Reprints or correspondence: Dr. Andrea De Luca, Institute of Clinical Infectious Diseases, Catholic University, Largo Francesco Vito, 1-00168 Rome, Italy (andrea.deluca{at}rm.unicatt.it)
 
Next Section

Abstract

To determine the variability of genotypic human immunodeficiency virus (HIV) type 1 drug-resistance interpretation by available expert systems and its clinical implications, 261 subjects for whom a potent antiretroviral regimen was failing who were starting salvage therapy were evaluated. The association of the genotypic susceptibility score (GSS) of the salvage regimen, according to 11 interpretation systems, with HIV RNA outcomes for 6 months was examined. GSS was highly variable, as determined by the different interpretation systems, and showed independent correlation with changes from baseline HIV RNA levels at 6 months with 5 systems—Stanford hivdb, GuideLines 3.0, Retrogram 1.4, HIVresistanceWeb, and São Paulo University. Most GSSs predicted virologic response in regimens containing stavudine, lamivudine, efavirenz, or indinavir. Selected systems predicted response in regimens containing didanosine, abacavir, or nelfinavir, and no system predicted outcome of boosted protease inhibitors. GSSs predicted changes in HIV RNA levels better in adherent patients than in nonadherent individuals. Interpretation may be improved, and knowledge should be used uniformly throughout different expert systems

The efficacy of highly active antiretroviral therapy (HAART) is strictly dependent on its ability to confer a potent and sustained control of human immunodeficiency virus type 1 (HIV-1) replication [1]. The development of HIV-1 drug resistance is a major determinant of failure to suppress virus load [2]. Several retrospective studies have established a significant association between baseline genotypic or phenotypic drug resistance and subsequent virologic response [3 –9]. Moreover, prospective studies have shown that the use of HIV-1 drug-resistance assays to guide treatment choice for patients who do not respond to HAART improves virologic outcomes [10 –15]

Several assays for determining HIV-1 drug susceptibility have been standardized and are now commercially available [16]. Genotypic assays are based on nucleic acid sequencing of the viral reverse-transcriptase and protease genes and identification of mutations associated with resistance to antiretroviral drugs, whereas phenotypic assays are based on in vitro measurement of susceptibility to individual drugs of a recombinant clone containing sequences from the tested virus, compared with that of a reference strain [17]. Genotypic assays have a faster turnaround time, are less technically complex, and are less expensive than phenotypic assays, which are usually performed only by centralized commercial facilities; therefore, genotypic assays are more widely used than phenotypic assays. Nevertheless, a major difficulty with resistance genotyping is the interpretation of results. Rough lists of mutations need some degree of interpretation by clinicians and virologists who are experts in the field before these lists can be used in clinical practice. Therefore, a large number of genotypic resistance interpretation tools have been developed in recent years [18]. These include mutation lists, rule-based algorithms, and interpretations based on databases correlating genotypes with corresponding phenotypic susceptibilities, all of which are available from a number of sources, including commercial sources [19 –26]. An initial comparative assessment of these interpretation tools [27, 28] showed some variability, which may modulate the choice of drugs to be used and therefore affect the therapeutic outcome

To analyze the variability of interpretation of HIV-1 genotypic drug resistance of available expert systems and to compare their ability to predict virologic and immunologic responses, we retrospectively analyzed the relationship between the different interpretations of resistance genotypes by several systems and the outcomes of salvage treatment in a cohort of HIV-1–infected patients who did not respond to potent combination antiretroviral therapy

Previous SectionNext Section

Patients and Methods

Study patients From among the HIV-1–infected patients followed at 2 HIV-referral centers (sites A and B) in Rome, we selected those who had received potent antiretroviral treatment for at least 3 months and who developed virologic failure between April 1999 and March 2001. Virologic failure was defined as the lack of a 1-log10 reduction of HIV-1 load after at least 3 months of therapy or virologic rebound after suppression of virus load, and, in both cases, an HIV RNA level >1000 copies/mL. We further selected patients who had a genotypic resistance test performed while still receiving the failing regimen, underwent a treatment change, and continued salvage treatment without major interruptions for the subsequent 6 months

Virologic and immunologic assays Genotyping was performed at the 2 study sites by use of 2 different commercial assays based on sequencing of cDNA derived from plasma HIV RNA: the TruGene assay (version 2.0; Visible Genetics) and the ViroSeq HIV-1 Genotyping System (Applied Biosystems). The former assay generates a 1.3-kb sequence from the pol region encompassing the entire protease gene and the major part of the reverse-transcriptase gene by bidirectional automated sequencing on the Microgene Clipper (Visible Genetics). The latter assay generates a 1.8-kb sequence from the same regions, using the ABI Prism automated sequencer 3700 (Applied Biosystems). The accuracy of all sequences was verified manually (M.G.R., A.B., and C.F.P.) in both assays

Drug-resistance results were used to guide treatment decisions for 76% of the patients. For these patients, results with ViroSeq were interpreted by an expert virologist (C.F.P.), and results with TruGene were interpreted by a previous interpretation table from Visible Genetics (vgi.gnl), with additional expert advice (A.D.L. and A.C.) [13]. None of the interpretation systems analyzed in this study was used for treatment decisions in this group of patients. For the remaining 24% of patients, drug-resistance genotyping was performed retrospectively, on stored plasma samples; therefore, results were not available during the treatment decision process

Plasma HIV RNA concentrations were measured by use of a bDNA assay with a detection limit of 50 copies/mL (Versant HIV RNA 3.0; Bayer Diagnostics). Peripheral blood CD4+ lymphocyte counts were performed by use of standard flow cytometry

Interpretation systems and genotypic susceptibility scores (GSSs) Baseline genotypic mutation patterns from each patient were reinterpreted by use of 11 different expert systems that consisted of rule-based algorithms, with the exception of the HIVresistanceWeb system, which consisted of tables of mutations. We used the most recent versions available as of 1 July 2001 (table 1)

Figure 1
View larger version:
Figure 1

Odds ratios (ORs) for human immunodeficiency virus RNA levels <500 copies/mL at 3 (A) and 6 (B) months after initiation of salvage treatment for each unit increase in the genotypic susceptibility score (GSS), according to each interpretation system. White squares and lines with whiskers indicate crude ORs and 95% confidence intervals (CIs); black squares with whiskers indicate adjusted ORs with 95% CIs. System sources are as follows: Stanford HIV Resistance database (hivdb; Stanford University, Stanford, CA); Agence Nationale de Recherches sur le SIDA (ANRS) AC11 (ANRS, Paris, France); Rega 4.0 (Katholieke Universiteit, Leuven, Belgium); GuideLines 3.0 (Visible Genetics, Toronto, Canada); Retrogram 1.4 (Virology Networks, Utrecht, The Netherlands); HIVresistanceWeb (Vertibrae, NY); Centre Hopitalier Luxemburg (CHL) version 3.2, Detroit Medical Center, and São Paulo University (Advanced Biological Laboratories, Luxemburg). RCG, Resistance Collaborative Group

Figure 2
View larger version:
Figure 2

Adjusted odds ratios (ORs) for human immunodeficiency virus RNA levels <500 copies/mL at 6 months (diamonds) for each unit increase in the genotypic susceptibility score derived from each interpretation system for salvage regimens containing abacavir (A; n=73), stavudine (B; n=195), didanosine (C; n=115); lamivudine (D; n=91), efavirenz (E; n=82), nelfinavir (F; n=90), indinavir (G; n=56), and pharmacologically boosted protease inhibitors (H; n=60). Bars indicate 95% confidence intervals (CIs). System sources are as follows: Stanford HIV Resistance database (hivdb; Stanford University, Stanford, CA); Agence Nationale de Recherches sur le SIDA (ANRS) AC11 (ANRS, Paris, France); Rega 4.0 (Katholieke Universiteit, Leuven, Belgium); GuideLines 3.0 (Visible Genetics, Toronto, Canada); Retrogram 1.4 (Virology Networks, Utrecht, The Netherlands); HIVresistanceWeb (Vertibrae, NY); Centre Hopitalier Luxemburg (CHL) version 3.2, Detroit Medical Center, and São Paulo University (Advanced Biological Laboratories, Luxemburg). RCG, Resistance Collaborative Group

Table 1
View larger version:
Table 1

Characteristics of analyzed interpretations systems for genotypic human immunodeficiency virus type 1 resistance and scores derived for individual drugs in the salvage regimens, according to systems interpretation

The interpretations provided by the different systems were translated into numeric values; each drug in the salvage regimen was assigned a susceptibility score ranging from 0 to 1, according to the criteria listed in table 2. With the exception of 3 algorithms [20, 21, 25], the interpretation systems did not themselves suggest a numeric drug susceptibility value. For this reason, we chose to translate definitions of “resistance” and “susceptible or no evidence of resistance” into susceptibility scores of 0 and 1, respectively. Definitions falling between these 2 extremes were arbitrarily translated into susceptibility values of 0.5, as described elsewhere [20, 25], except when the authors of the systems either suggested different numeric values [21] or when there were 2 clearly distinct intermediate susceptibility values (Retrogram 1.4). In the latter case, 2 equally distant intermediate values were chosen (table 1). The sum of the scores of the individual drugs in the salvage regimen provided the GSS of that regimen. Ritonavir, used at doses ⩽200 mg every ⩽12 h, was not considered to be an active drug. For regimens in which protease inhibitors (PIs; saquinavir, amprenavir, or indinavir) were pharmacokinetically boosted with ritonavir, interpretations and, consequently, scores were identical to those of the corresponding nonboosted regimens, except when specified by the system

Table 2
View larger version:
Table 2

Genotypic susceptibility scores (GSSs) of the salvage regimens of study patients, according to each system

Adherence Patient-reported adherence to antiretroviral treatment was assessed in a subgroup of individuals by use of a self-administered, previously validated questionnaire, as described elsewhere [31]. Patients were categorized in 2 major groups: adherent and nonadherent, as described elsewhere [13]

Statistical analysis All HIV RNA values were log-transformed before analysis. Continuous variables were compared by Student’s t test, and categorical variables were compared by the χ2 test. The t test for dependent samples was used to compare GSSs given to patients’ salvage treatment by the different interpretation systems. Pearson’s parametric test was used to analyze the correlation among the different scores. The associations of variables with dichotomous end points were analyzed by bivariate and multivariable logistic regression models. The associations with continuous outcomes were analyzed by simple and multiple linear regression models. Analyses were performed by use of the SPSS statistical software package (version 9.0)

Previous SectionNext Section

Results

Patient baseline characteristics A total of 261 patients fulfilling the selection criteria were included in the study: 122 from site A and 139 from site B. The median age was 38 years (range, 18–69 years), and 29% were female. HIV transmission categories were as follows: male homosexual (26%), injection drug use (26%), heterosexual (38%), blood transfusion (2%), and unknown (8%). Thirty-seven percent had a previous diagnosis of AIDS (Centers for Disease Control and Prevention class C criteria [32]). The median HIV RNA level was 4.52 log10 copies/mL (range, 3.00–6.18 log10 copies/mL), and the median CD4+ cell count was 287 cells/μL (range, 5–1100 cells/μL). Before genotyping, patients had been receiving a potent combination regimen for a median of 24 months (range, 4–85 months) and nucleoside reverse-transcriptase inhibitors (NRTIs) for a median of 35 months (range, 4–131 months); they had had virologic failure with a median of 2 potent regimens (range, 1–9 regimens). Thirty-nine percent of patients had been exposed to 3 antiretroviral drug classes. Eighty-three patients (32%) reached, at least once, an HIV RNA level <500 copies/mL during previous treatment. Comparison between centers yielded a significantly higher baseline HIV RNA level (mean±SD, 4.7±0.6 vs. 4.4±0.6 log10 copies/mL), a longer period of time receiving a potent regimen (mean±SD, 28.6±11.3 vs. 20.5±7.8 months), and a higher number of previously failed regimens (mean±SD, 2.9±1.8 vs. 2.3±1.2) in patients from site A versus those from site B. The other characteristics did not differ between study sites

Baseline genotyping results and GSSs The genotyping results showed a median number of 9 overall resistance mutations (range, 2–17 resistance mutations) [33]. There was a median of 3 NRTI-resistance mutations (range, 0–8 NRTI-resistance mutations); those present in >10% of isolates were T215Y/F (59%), M184V (55.9%), M41L (49%), D67N (38.5%), L210W (34.5%), K70R (24.5%), K219Q/E (24.5%), and T69D (11.5%). Substitutions associated with resistance to nonnucleoside reverse-transcriptase inhibitors (NNRTIs) were detected in 40% of individuals; the most frequent NNRTIs wereK103N (19.9%), G190A/S (12.6%), and Y181C/I (11.5%). There was a median of 5 PI-resistance mutations (range, 0–9 PI-resistance mutations); those present in >10% of isolates were L63P (73.6%), L10F/I/R/V (48.3%), A71T/V (47.9%), L90M (44%), M46I/L (33.7%), V82A (33.3%), V77I (33%), M36I (32.2%), I54L/V (24.1%), I84V (17.2%), and G73A/S (13.4%). There was no difference in the number of mutations between study sites

Table 2 shows the mean GSSs of the salvage regimens generated with each system. Pairwise comparison of the mean GSSs showed that scores for the Stanford hivdb, GuideLines 3.0, and Menéndez-Arias algorithms differed significantly from each other and from the scores of each of the other systems. Other systems generated mean GSSs that differed significantly from some, but not all, systems (table 2). Despite the significantly different mean GSSs, parametric linear correlation showed that GSSs calculated using all the different systems were significantly correlated with one another (data not shown)

Salvage treatments Treatments administered after baseline consisted of 3 antiretroviral drugs in 75% and 4 or 5 drugs in 25% of patients; 36% of patients started a new drug class, and the median number of new drugs used (i.e., antiretroviral agents never used during previous treatment) was 2 (range, 0–4). Abacavir was given to 73 patients; didanosine, to 115; lamivudine, to 91; stavudine, to 195; zidovudine, to 30; efavirenz, to 82; nevirapine, to 47; amprenavir, to 31 (boosted by ritonavir, to 25); indinavir, to 56 (boosted by ritonavir, to 18); lopinavir/ritonavir, to 21; nelfinavir, to 90; ritonavir at therapeutic doses, to 22; and saquinavir hard-gel capsules, to 20 (boosted by ritonavir, to 9). Delavirdine, tenofovir, and zalcitabine were not used in the study patients. Patients from site A, compared with those from site B, were prescribed stavudine more frequently (85.2% vs. 65.4%; P=.0002) and efavirenz (45.1% vs. 19.4%; P<.0001) and zidovudine (4.9% vs. 17.3%; P=.002) and nelfinavir (35.4% vs. 42.4%; P=.004) less frequently

Virologic and immunologic outcomes The proportions of patients with an HIV RNA level <500 copies/mL after 3 and 6 months were 34% and 27%, respectively. The median change from baseline HIV RNA was −0.88 log10 copies/mL after 3 months and −0.52 log10 copies/mL after 6 months. The median baseline CD4+ T cell count change was +26 cells/μL after 3 months and +16 cells/μL after 6 months

Prediction of virologic outcomes Crude and adjusted odds ratios (ORs) for reaching virologic responses defined as plasma HIV RNA levels <500 copies/mL for each unit increase of the GSSs resulting from the different interpretation systems are reported in figure 1. All analyzed interpretation systems were significantly predictive of the virologic response, both after 3 and 6 months on bivariate analysis, with ORs ranging from 1.35 (São Paulo University) to 2.04 (Agence Nationale de la Recherches sur le SIDA [ANRS] AC11) at 3 months and from 1.44 (São Paulo University) to 2.10 (GuideLines 3.0) at 6 months for each unit increase of the GSS. The multivariable model included characteristics of patient history (previous AIDS diagnosis, previous virologic suppression with potent combination therapy, and number of previous potent regimens), baseline features (HIV RNA levels and number of resistance mutations), and characteristics of the salvage regimen (number of new and of total number of antiretroviral agents used and use of a new drug class), which are relevant in terms of salvage treatment decisions. Only 3 of 11 interpretation systems showed significant adjusted prediction of the 3-month response (ANRS AC11 [P=.005], Rega 4.0 [P=.041], and GuideLines 3.0 [P=.040]), and 4 of 11 showed significant adjusted prediction of the 6-month response (Stanford hivdb [P=.024], GuideLines 3.0 [P=.040], Retrogram 1.4 [P=.010], and HIVresistanceWeb [P=.020]) (figure 2). A separate analysis that also was adjusted by study site yielded almost identical results (data not shown)

We then used multiple linear regression models in which the GSSs given by the different interpretation systems were adjusted for by the same variables used in the multivariable logistic regression and were related to the logarithmic HIV RNA level change from baseline at 3 and 6 months (table 3). Models’ R 2 values (i.e., the proportion of virus load change at given times, explained by the variables in the models), using the different GSSs, ranged from 0.18 to 0.22 for 3-month changes and from 0.16 to 0.19 for 6-month changes. Two of 11 interpretation systems predicted significant virus load changes at 3 months (GuideLines 3.0 and Centre Hopitalier Luxemburg 3.2), and 5 of 11 predicted significant changes at 6 months (Stanford hivdb, GuideLines 3.0, Retrogram 1.4, HIVresistanceWeb, and São Paulo University)

Table 3
View larger version:
Table 3

Association of baseline genotypic susceptibility score (GSS) with change from baseline human immunodeficiency virus (HIV) RNA levels (multiple linear regression)

Role of patient-reported adherence Detailed patient-reported adherence data were available for a subgroup of 91 individuals from a single center. In adherent patients (n=53), GSSs from 9 of 11 interpretation systems (all except RCG and Menéndez-Arias algorithms) showed significant linear associations with virus load changes at 6 months. On the other hand, among patients who reported nonadherence (n=38), GSSs from only 1 of 11 interpretation systems (Centre Hopitalier Luxemburg 3.2) correlated with virologic responses (detailed results are available from authors on request)

Analysis of the interpretation of resistance for specific treatment groups ORs for virologic response at 6 months for treatment regimens containing specific agents given to >50 patients in the cohort were analyzed (see Methods). Multiple logistic regression models were used, in which GSSs derived from the different interpretation systems were adjusted by baseline HIV RNA levels and the number of new drugs in the salvage regimen [21]. The results are illustrated in figure 2. For abacavir-containing regimens, 2 of 11 interpretation systems were significantly predictive of virologic response (Stanford hivdb [P=.034] and Rega 4.0 [P=.019]), and 2 others were borderline predictive (ANRS AC11 [P=.078] and GuideLines 3.0 [P=.082]). All but 2 interpretation systems (Resistance Collaborative Group and Menéndez-Arias algorithms) predicted virologic responses for stavudine-containing regimens. For didanosine-containing regimens, 7 of 11 systems predicted significant virologic responses at 6 months (ANRS AC11, Rega 4.0, GuideLines 3.0, HIVresistanceWeb, Stanford hivdb, Retrogram 1.4, and Centre Hopitalier Luxemburg 3.2). Responses to lamivudine-containing regimens were predicted by 8 systems. Responses to efavirenz-containing regimens were significantly predicted by 9 of 11 interpretation systems. Only 3 systems showed significant prediction of responses to nelfinavir-containing regimens (ANRS AC11, GuideLines 3.0, and Retrogram 1.4). Seven of 11 interpretation systems could predict virologic responses to indinavir-containing regimens. Virologic responses to boosted PI regimens were not predicted by any GSS

Prediction of immunologic outcomes The association between GSSs and change from baseline CD4+ cell count after 3 and 6 months was analyzed by univariate linear regression; results are reported in table 4. Three of 11 interpretation systems predicted significant CD4+ cell count changes after 3 months, whereas 7 of 11 predicted significant CD4+ cell count changes after 6 months. After 6 months, the mean change from baseline CD4+ T cell count for each unit increase of GSS given by the different systems ranged from +15 to +40 cells/μL

Table 4
View larger version:
Table 4

Association of baseline genotypic susceptibility score (GSS) with change from baseline CD4+ cell counts (simple linear regression)

Previous SectionNext Section

Discussion

HIV-1 drug resistance is an important factor responsible for failure of antiretroviral treatment. Genotyping of the pol region of HIV-1 by population nucleic acid sequencing is the most commonly used method for determining viral resistance. Nevertheless, the interpretation of the clinical relevance of different associations of amino acid substitutions in patients receiving combination therapy remains a major challenge and requires expert knowledge [12, 34]. To help clinicians interpret the relevance of the mutational pattern of patients’ virus, several expert systems have been developed. In this study, we observed that, after translating interpretations of baseline genotypes by 11 different systems in numerical values, the salvage regimens of a cohort of 261 patients were assigned heterogeneous susceptibility scores. Similarly, other authors found significant variation in the interpretation of genotypic resistance by several algorithms; in particular, the interpretation of resistance to selected NRTIs was found to differ more frequently [27, 35]

The clinical relevance of these different interpretations was examined by retrospectively correlating sensitivity scores derived from the interpretation of drug resistance given by the different systems to the virologic response to the salvage regimen. Multivariable logistic regression models were used to adjust the prediction of the different interpretation systems for clinical variables, baseline virus load, and number of resistance mutations, which represents a raw result of the genotyping assay without any interpretation, as well as the total number and type of new drugs used in the salvage regimen. Results showed that only 4 of the 11 interpretation systems predicted 6-month virologic success. This means that genotype interpretation by these systems had an adjunctive predictive value over the cited clinical variables and the raw genotyping result. Similar results were obtained by use of multiple linear regression models that adjusted for the same factors and related the sensitivity scores to the logarithmic change from baseline HIV RNA levels; scores from 5 of 11 interpretation systems were independently predictive of 6-month virus load outcomes. These results show for the first time that discordant interpretation may influence the predictive value of resistance testing over subsequent virologic outcomes. In general, we observed that algorithms that were updated in 2001 performed better than algorithms updated the previous year, which is consistent with a positive knowledge evolution in the genotypic resistance interpretation field. Clearly, a definitive comparative judgment about the clinical utility of different interpretation tools requires randomized, controlled clinical trials, although the speedy development of interpretation knowledge and the consequent rapid updating of existing systems make such studies difficult to implement

We then attempted to establish the predictive value of genotypic resistance interpretation to individual agents, analyzing the virologic response to regimens containing specific drugs that were used in a significant number of cases in the cohort. Because of the more limited number of outcomes, simpler logistic regression models were used for these analyses, which were adjusted for baseline virus load and number of new drugs used in the salvage treatment, as described elsewhere [21]. Interestingly, although most interpretation systems gave significant independent prediction of success in stavudine-containing regimens and in lamivudine-containing regimens, for didanosine- or abacavir-containing regimens, a minority of rule-based algorithms predicted virologic success at both end points. These results might reflect the relatively better knowledge of genotypic correlates of clinically relevant resistance to stavudine and lamivudine relatively to abacavir and didanosine [36]

The present study was not designed to evaluate the predictive value of individual rules of the interpretation systems. Indeed, rules were uncovered only for a part of the systems, and their evaluation was influenced by the prevalence of specific mutations and combination of mutations in the cohort, as well as by the strength of the association of the interpretation rules for the other drugs in the combination regimen with treatment outcomes. Nonetheless, some rough indications can be deduced from the results. The algorithm from Menéndez-Arias was developed before evidence that nucleoside analogue mutations (NAMs) played a determinant role in NRTI cross-resistance was available [34]; therefore, this algorithm did not consider these mutations for interpreting resistance to stavudine, didanosine, and abacavir. This, together with the high prevalence of NAMs in the cohort, is at least in part responsible for the poor association of the interpretation provided by the algorithm with treatment response in regimens containing these nucleoside analogues. The same occurred with the algorithm from the Resistance Collaborative Group, which gave minor consideration to NAMs in the interpretation of resistance to didanosine and stavudine. The HIVresistanceWeb mutation tables did not take into account the possible role of the incremental number of mutations, including NAMs, for determining resistance to abacavir and showed no association between GSS and treatment response in abacavir-containing regimens. On the other hand, association with outcomes in regimens containing stavudine, didanosine, and abacavir was better by use of other rule-based algorithms, such as Stanford hivdb, ANRS AC11, Rega 4.0, GuideLines 3.0, and Retrogram 1.4. The first 4 of these systems incorporated NAMs in the resistance rules for these drugs; the rules of the latter algorithm were not available for evaluation. Altogether, these findings highlight the importance of this group of mutations for determining clinical drug resistance and cross-resistance to several nucleoside analogues

The GSSs of efavirenz- and of indinavir-containing salvage regimens were significantly associated with virological outcomes, as determined by use of most interpretation systems. In the first case, this probably reflects the simple mechanisms of genotypic resistance to NNRTIs, which are often based on the presence of single mutations conferring high levels of phenotypic resistance; this also was confirmed by the significant association of scores for nevirapine-based regimens with outcomes (data not shown) [37]. In the case of indinavir, homogeneous correct interpretation was probably related to the established knowledge of the phenotypic correlates of genotypic resistance and good correlation between phenotypes and clinical response [38]. On the other hand, sensitivity scores of nelfinavir-based regimens and PI-based salvage therapy with pharmacokinetic enhancement were rarely, if ever, associated with virologic outcomes. This lower predictive value might have been in part influenced by the low potency of nelfinavir in the context of salvage therapy or might suggest that the correct understanding of resistance and cross-resistance to nelfinavir requires more investigation and that clinically relevant rules needed to be incorporated by some of the algorithms. Finally, the results indicate that the role of pharmacokinetic enhancement of PIs in overcoming partial resistance to some agents of this class, as well as the genotypic correlates of resistance to boosted regimens, need more exploration [9]

Overall, these findings indicate that the understanding of genotypic HIV-1 drug resistance has recently improved but also that further improvement is necessary. That several interpretation systems that were analyzed in this study are regularly updated and that the versions presented here may already be superseded must be considered

Drug resistance is not the only reason for virologic failure of HAART regimens. Another major determinant is incomplete medication adherence [39 –41]. In a subanalysis, we observed that patients who reported nonadherence markedly reduced the association between baseline drug susceptibility, as interpreted by the different systems, and virologic outcomes of salvage therapy. This suggests that the optimal understanding of drug resistance from data from clinical outcomes requires appropriate control of other potentially confounding variables. Although we adjusted for interpretation by major clinical and virologic confounders, we could not adjust for adherence, because detailed analysis was available only for a subset of patients. Therefore, a proportion of nonadherent patients might have diluted the overall strength of the association between the different interpretations of drug resistance and the subsequent responses, and outcome prediction by GSSs might have been underestimated. On the other hand, patients in this cohort probably reflect “real world” findings and might be more directly comparable to clinical practice

Another possible limitation of this study lies in the translation of the indications from the interpretation systems into numeric values. With the exception of 3 interpretation systems, in which the numeric translations were taken directly from those given by their authors, this was made arbitrarily. Nevertheless, the translation process adhered to the same principles used by authors of HIV drug-resistance algorithms and used homogeneous criteria for all the systems [20, 21, 25]. Nonetheless, assignment of numeric values different from those chosen in this study, especially for the translation of intermediate values of resistance, might have influenced the predictive value of some of the systems

Finally, all drugs in the salvage regimen were given the same weight in the scoring system, without taking drug potency into account. Although this also represents a limitation, previous analyses correlating GSSs or phenotypic susceptibility scores to virologic outcomes as well as other approaches with similar concepts, such as the number of active drugs, always were used without adjustment for drug potency, with significant results [11, 14, 21, 25, 42, 43]. Furthermore, assigning different values according to drug potency would have introduced other arbitrary correction factors. In fact, there is no clear in vivo reference evaluation of the potency of one antiretroviral agent or drug class versus that of another

Virologic outcomes are the major benchmark for testing the real clinical value of genotypic resistance interpretation rules and for constructing new rules for available and investigational antiretroviral drugs [44 –46]. For this reason, there is a need for wide-ranging databases containing appropriately quality-controlled data from genotypic resistance assays, treatments, and clinical outcomes. Correlation among mutational patterns, specific treatments, and virologic outcomes might be explored in such databases by use of standard statistics, as well as newer data-mining approaches [47, 48]

In conclusion, we found significant interpretation discordance among several available systems for interpreting HIV-1 genotypic resistance. These heterogeneous interpretations were associated with different predictions of subsequent virologic outcomes. There is an urgent need for a joint effort to develop, validate, and publish standardized rules and definition criteria for HIV-1 genotypic resistance interpretation, to provide accessible tools for interpretation that help improve the clinical usefulness of genotypic assay results

Previous SectionNext Section

Acknowledgments

We are grateful to Alessandra Bacarelli, for technical assistance; Wilco Keulen (Virology Networks) and Advanced Biological Laboratories, for providing free access to their interpretation systems; and Visible Genetics, for providing access to the rules of their interpretation system GuideLines

Previous SectionNext Section

Footnotes

  • Presented in part: 3rd European Symposium on the Clinical Implications of HIV Drug Resistance, Frankfurt, Germany, 23–25 February 2001 (abstract 37); 5th International Workshop on HIV Drug Resistance and Treatment Strategies, Scottsdale, Arizona, 5–9 June 2001 (abstract 91)

    Informed consent was obtained from subjects in accordance with guidelines of the local institutions where the study was conducted

    Financial support: Ministero della Sanità, Istituto Superiore di Sanità, III e IV Programma Nazionale di Ricerca sull’AIDS “Patologia, Clinica e Terapia dell’Infezione da HIV”; Ricerca Corrente e Finalizzata degli Istituto di Ricovero e Cura a Carattere Scientifico

  • Received November 29, 2002.
  • Accepted January 7, 2003.
Previous Section
 

References

  1. ↵
    1. Demeter LM,
    2. Hughes MD,
    3. Coombs RW,
    4. et al
    . Predictors of virologic and clinical outcomes in HIV-1–infected patients receiving concurrent treatment with indinavir, zidovudine and lamivudine. AIDS Clinical Trials Group Protocol 320. Ann Intern Med 2001;135:954-64.
    Abstract/FREE Full Text
  2. ↵
    1. Hirsch MS,
    2. Brun-Vezinet F,
    3. D’Aquila RT,
    4. et al
    . Antiretroviral drug resistance testing in adult HIV-1 infection: recommendations of an International AIDS Society–USA Panel. JAMA 2000;283:2417-26.
    Abstract/FREE Full Text
  3. ↵
    1. Falloon J,
    2. Ait-Khaled M,
    3. Thomas DA,
    4. et al
    . HIV-1 genotype and phenotype correlate with virological response to abacavir, amprenavir and efavirenz in treatment-experienced patients. AIDS 2002;16:387-96.
    CrossRefMedlineWeb of Science
  4. ↵
    1. Servais J,
    2. Plesseria JM,
    3. Lambert C,
    4. et al
    . Genotypic correlates of resistance to HIV-1 protease inhibitors on longitudinal data: the role of secondary mutations. Antivir Ther 2001;6:239-48.
    MedlineWeb of Science
  5. ↵
    1. Lorenzi P,
    2. Opravil M,
    3. Hirschel B,
    4. et al
    . Impact of drug resistance mutations on virologic response to salvage therapy. Swiss HIV Cohort Study. AIDS 1999;13:F17-21.
    CrossRefMedlineWeb of Science
  6. ↵
    1. Kuritzkes DR,
    2. Sevin A,
    3. Young B,
    4. et al
    . Effect of zidovudine resistance mutations on virologic response to treatment with zidovudine-lamivudine-ritonavir: genotypic analysis of human immunodeficiency virus type 1 isolates from AIDS Clinical Trials Group protocol 315. ACTG Protocol 315 Team. J Infect Dis 2000;181:491-7.
    Abstract/FREE Full Text
  7. ↵
    1. Lawrence J,
    2. Schapiro J,
    3. Winters M,
    4. et al
    . Clinical resistance patterns and responses to two sequential protease inhibitor regimens in saquinavir- and reverse transcriptase inhibitor–experienced persons. J Infect Dis 1999;179:1356-64.
    Abstract/FREE Full Text
  8. ↵
    1. Harrigan PR,
    2. Montaner JS,
    3. Hogg RS,
    4. et al
    . Baseline HIV drug resistance profile predicts response to ritonavir/saquinavir protease inhibitor therapy in a community setting. AIDS 1999;13:1863-71.
    CrossRefMedlineWeb of Science
  9. ↵
    1. Zolopa AR,
    2. Shafer RW,
    3. Warford A,
    4. et al
    . HIV-1 genotypic resistance patterns predict response to saquinavir-ritonavir therapy in patients in whom previous protease inhibitor therapy had failed. Ann Intern Med 1999;131:813-21.
    Abstract/FREE Full Text
  10. ↵
    1. Durant J,
    2. Clevenbergh P,
    3. Halfon P,
    4. et al
    . Drug-resistance genotyping in HIV-1 therapy: the VIRADAPT randomised controlled study. Lancet 1999;353:2195-9.
    CrossRefMedlineWeb of Science
  11. ↵
    1. Baxter JD,
    2. Mayers BL,
    3. Wentworth DN,
    4. et al
    . A randomized study of antiretroviral management based on plasma genotypic resistance testing in patients failing therapy. AIDS 2000;14:F83-F93.
    CrossRefMedlineWeb of Science
  12. ↵
    1. Tural C,
    2. Ruiz L,
    3. Holtzer C,
    4. et al
    . Clinical utility of HIV-1 genotyping and expert advice: the Havana trial. AIDS 2002;16:209-18.
    CrossRefMedlineWeb of Science
  13. ↵
    1. Cingolani A,
    2. Antinori A,
    3. Rizzo MG,
    4. et al
    . Usefulness of monitoring HIV drug resistance and adherence in individuals failing highly active antiretroviral therapy: a randomized study (ARGENTA). AIDS 2002;16:369-79.
    CrossRefMedlineWeb of Science
  14. ↵
    1. Cohen CJ,
    2. Hunt S,
    3. Sension M,
    4. et al
    . A randomized trial assessing the impact of phenotypic resistance testing on antiretroviral therapy. AIDS 2002;16:579-88.
    CrossRefMedlineWeb of Science
  15. ↵
    1. Meynard JL,
    2. Vray M,
    3. Morand-Joubert L,
    4. et al
    . Phenotypic or genotypic resistance testing for choosing antiretroviral therapy after treatment failure: a randomized trial. AIDS 2002;16:727-36.
    CrossRefMedlineWeb of Science
  16. ↵
    1. The EuroGuidelines Group for HIV Resistance
    . Clinical and laboratory guidelines for the use of HIV-1 drug-resistance testing as part of treatment management: recommendations for the European setting. AIDS 2001;15:309-20.
    CrossRefMedlineWeb of Science
  17. ↵
    1. Hertogs K,
    2. de Bethune MP,
    3. Miller V,
    4. et al
    . A rapid method for simultaneous detection of phenotipic resistance to inhibitors of protease and reverse transcriptase in recombinant human immunodeficiency type 1 isolates from patients treated with antiretroviral drugs. Antimicrob Agents Chemother 1998;42:269-76.
    Abstract/FREE Full Text
  18. ↵
    1. Schapiro JM,
    2. De Luca A,
    3. Harrigan R,
    4. et al
    . Resistance assay interpretation systems vary widely in method and approach [abstract 172]. Antivir Ther 2001;6((Suppl 1)):131.
  19. ↵
    1. Shafer RW,
    2. Jung DR,
    3. Betts BJ,
    4. Xi Y,
    5. Gonzales MJ
    . Human immunodeficiency virus reverse transcriptase and protease sequence database. Nucleic Acids Res 2000;28:346-8.
    Abstract/FREE Full Text
  20. ↵
    1. Stanford University
    . Stanford HIV Drug Resistance Database. Available at: http://hivdb.stanford.edu. Accessed 26 November. 2002.
  21. ↵
    1. DeGruttola V,
    2. Dix L,
    3. D’Aquila R,
    4. Aquila R,
    5. 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. Antivir Ther 2000;5:41-8.
    MedlineWeb of Science
  22. ↵
    1. Larder BA,
    2. Kemp SD,
    3. Hertogs K
    . Quantitative prediction of HIV-1 phenotypic drug resistance from genotypes: the virtual phenotype (Virtual Phenotype). Antivir Ther 2000;5((Suppl 3)):49-50.
    MedlineWeb of Science
  23. ↵
    1. Clotet B,
    2. Menéndez-Arias L,
    3. Ruiz L,
    4. et al.
    1. Menéndez-Arias L,
    2. Domingo E
    . Algorithms for the clinical use of genotypic information. In: Clotet B, Menéndez-Arias L, Ruiz L, et al., editors. Guide to management of HIV resistance and pharmacokinetics of drug therapy. 1st ed. Barcelona: Taisa editorial; 2000. p. 83.
  24. ↵
    1. Clotet B,
    2. Menéndez-Arias L,
    3. Ruiz L,
    4. et al.
    1. Vandamme A-M
    . Possible genotypic and phenotypic algorithm for the clinical use of HIV drug resistance testing. In: Clotet B, Menéndez-Arias L, Ruiz L, et al., editors. Guide to management of HIV resistance and pharmacokinetics of drug therapy. 1st ed. Barcelona: Taisa editorial; 2000. p. 85-9.
  25. ↵
    1. van Laethem K,
    2. De Luca A,
    3. Antinori A,
    4. Cingolani A,
    5. Perno CF
    . A genotypic drug resistance algorithm that significantly predicts therapy response in HIV-1–infected patients. Antivir Ther 2002;7:123-9.
    MedlineWeb of Science
  26. ↵
    1. Boulmé R,
    2. Halfon P,
    3. Diaz R,
    4. Dugas O,
    5. Schmit JC
    . Comparative evaluation between five automated resistance interpretation algorithms [abstract 156]. Antivir Ther 2001;6((Suppl 1)):121.
  27. ↵
    1. Shafer RW,
    2. Gonzales MJ,
    3. Brun-Vézinet F
    . Online comparison of HIV-1 drug resistance algorithms identifies rates and causes of discordant interpretations [abstract 134]. Antivir Ther 2001;6((Suppl 1)):101.
  28. ↵
    Guidelines from the Agence Nationale de Recherches sur le. SIDA. Available at: http://www.sante.gouv.fr/htm/actu/rap_delf/4_genotype.htm/. Accessed 26 November. 2002.
    1. Vertibrae
    . HIVresistanceWeb. Available at: http://www.hivresistanceweb.com/protected/mutationtables/. Accessed 26 November. 2002.
    1. Advanced Biological Laboratories
    . Available at: http://www.ablnetworks.org. Accessed 26 November. 2002.
  31. ↵
    1. Murri R,
    2. Ammassari A,
    3. Gallicano K,
    4. et al
    . Patient-reported nonadherence to HAART is related to protease inhibitors levels. J Acquir Immune Defic Syndr 2000;24:123-8.
    MedlineWeb of Science
  32. ↵
    1. Centers for Disease Control and Prevention
    . 1993 Revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Morb Mortal Wkly Rep 1992;41((RR-17)):1-19.
    Medline
  33. ↵
    1. International AIDS Society–USA Drug Resistance Mutations Group
    . Update on drug resistance mutations in HIV-1. Top HIV Med 2001;9:21-3.
  34. ↵
    1. Shafer RW
    . Genotypic testing for human immunodeficiency virus type 1 drug resistance. Clin Microbiol Rev 2002;15:247-77.
    Abstract/FREE Full Text
  35. ↵
    1. Schmidt B,
    2. Walter H,
    3. Schwingel E,
    4. et al
    . Comparison of different interpretation systems for genotypic HIV-1 drug resistance data [abstract 135]. Antivir Ther 2001;6((Suppl 1)):102.
  36. ↵
    1. Walter H,
    2. Schmidt B,
    3. Werwein E,
    4. Schwingel E,
    5. Korn K
    . Prediction of abacavir resistance from genotypic data: impact of zidovudine and lamivudine resistance in vitro and in vivo. Antimicrob Agents Chemother 2002;46:89-94.
    Abstract/FREE Full Text
  37. ↵
    1. Richman DD,
    2. Havlir D,
    3. Corbeil J,
    4. et al
    . Nevirapine resistance mutations of human immunodeficiency virus type 1 selected during therapy. J Virol 1994;68:1660-6.
    Abstract/FREE Full Text
  38. ↵
    1. Condra JH,
    2. Holder DJ,
    3. Schleif WA,
    4. et al
    . Genetic correlates of in vivo viral resistance to indinavir, a human immunodeficiency virus type 1 protease inhibitor. J Virol 1996;70:8270-6.
    Abstract/FREE Full Text
  39. ↵
    1. Haubrich RH,
    2. Little SJ,
    3. Currier JS,
    4. et al
    . The value of patient-reported adherence to antiretroviral therapy in predicting virologic and immunologic response. California Collaborative Treatment Group. AIDS 1999;13:1099-107.
    CrossRefMedlineWeb of Science
  40. ↵
    1. Bangsberg DR,
    2. Hecht FM,
    3. Charlebois ED,
    4. et al
    . Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population. AIDS 2000;14:357-66.
    CrossRefMedlineWeb of Science
  41. ↵
    1. Mannheimer S,
    2. Riedland G,
    3. Matts J,
    4. Child C,
    5. Chesney M
    . The consistency of adherence to antiretroviral therapy predicts biologic outcomes for human immunodeficiency virus–infected persons in clinical trials. Clin Infect Dis 2002;34:1115-21.
    Abstract/FREE Full Text
  42. ↵
    1. Baxter JS,
    2. Merigan TC,
    3. Wentworth DN,
    4. et al
    . Both baseline HIV-1 drug resistance and antiretroviral drug levels are associated with short-term virologic responses to salvage therapy. AIDS 2002;16:1131-8.
    CrossRefMedlineWeb of Science
  43. ↵
    1. Katzenstein D,
    2. Bosch RJ,
    3. Wang N,
    4. Hellmann NS,
    5. Bacheler L
    . Program and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections. Alexandria, VA: Foundation for Retrovirology and Human Health; 2002. Baseline phenotypic susceptibility and virologic failure over 144 weeks among nucleoside RT inhibitor–experienced subjects in ACTG 364 [abstract 591]. ACTG 364 Study Team.
  44. ↵
    1. Lanier ER,
    2. Scott J,
    3. Ait-Khaled M,
    4. et al
    . Predicting abacavir antiviral activity using HIV-1 reverse transcriptase genotype: a comparison of 12 algorithms [abstract 137]. Antivir Ther 2001;6((Suppl 1)):103.
  45. ↵
    1. Isaacson J,
    2. Kempf D,
    3. Calvez V,
    4. et al
    . Program and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections. Alexandria, VA: Foundation for Retrovirology and Human Health; 2002. Quantitative estimate of the effect of baseline mutations in HIV protease on the virologic response to lopinavir/ritonavir therapy in heavily antiretroviral-experienced patients [abstract 559].
  46. ↵
    1. Miller MD,
    2. Margot NA,
    3. Lu B
    . Program and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections. Alexandria, VA: Foundation for Retrovirology and Human Health; 2002. Effect of baseline nucleoside-associated resistance on response to tenofovir DF therapy: integrated analyses of studies 902 and 907 [abstract 43].
  47. ↵
    1. Descamps D,
    2. Masquelier B,
    3. Mamet JP,
    4. et al
    . A genotypic sensitivity score for amprenavir based on genotype at baseline and virological response [abstract 136]. Antivir Ther 2001;6 Suppl 1:103-4.
  48. ↵
    1. Beerenwinkel N,
    2. Schmidt B,
    3. Walter H,
    4. et al
    . Diversity and complexity of HIV-1 drug resistance: a bioinformatics approach to predicting phenotype from genotype. Proc Natl Acad Sci USA 2002;99:8271-6.
    Abstract/FREE Full Text
| Table of Contents

This Article

  1. J Infect Dis. (2003) 187 (12): 1934-1943. doi: 10.1086/375355
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 205 (5)
  1. Journal of Infectious Diseases
  1. Alert me to new issues

Most