Genotypic and phenotypic resistance patterns at virological failure in a simplification trial with nevirapine, efavirenz or abacavir

Background: The NEFA Study was a randomized study comparing nevirapine (NVP), efavirenz (EFV) or abacavir (ABC) as substitutes for protease inhibitors in a large group of HIV-1-infected patients successfully treated with antiretroviral regimens containing protease inhibitors.

Objective: To evaluate genotype and phenotype resistance patterns among patients who have experienced virological failure under one of the three study arms.

Methods: Patients with virological failure, defined as two consecutive determinations of HIV-1 RNA > 200 copies/ml, were analysed for phenotypic susceptibility and HIV-1 mutations.

Results: Of the 460 patients included in the study, 51 (11%) experienced virological failure after 24 months of follow-up while on assigned study medication. A higher proportion of patients in the ABC [25 (17%)] than in the NVP [14 (9%)] or EFV [12 (8%)] arms selected resistance to the study drug (P = 0.04). Moreover, a much higher number of resistance mutations to one or more of the backbone nucleoside reverse transcriptase inhibitor drugs contained in the failing regimen were observed in the ABC than in the EFV or NVP arms. In general, there was a good concordance among genotype and phenotype resistance testing, except for ABC, stavudine and didanosine, where phenotypic resistance testing added valuable information (fold change in the median inhibitory concentration).

Conclusions: Cross-resistance involving nucleoside reverse transcriptase inhibitor drugs might explain the higher risk of virological failure in patients switched to ABC-containing antiretroviral therapy. Phenotypic resistance testing may be helpful in interpreting unclear genotypic results.

Introduction

The combination of a protease inhibitor (PI) plus two nucleoside analogue reverse transcriptase inhibitors (NRTI) in highly active antiretroviral therapy (HAART) is complex because it involves a high daily pill burden and it has been associated with side effects potentially leading to lack of adherence and consequently to a virological failure, with or without selection of resistant mutants ^[1,2].

Assuming that simplified regimens with PI-sparing combinations in patients with virological suppression could be easier to use but equally potent, several studies have been performed, mainly with abacavir (ABC) ^[3,4] and non-nucleoside reverse transcriptase inhibitors (NNRTI) ^[5,6]. The Nevirapine, Efavirenz and Abacavir (NEFA) Study was a randomized, open-label, multicentre trial comparing three different simplified antiretroviral strategies with nevirapine (NVP), efavirenz (EFV) or ABC as substitutes for a PI drug in a large group of HIV-1-infected patients with plasma HIV-1 RNA levels < 200 copies/ml for longer than 6 months and receiving combination therapy with two NRTI plus a PI. The main study demonstrated a trend towards a higher virological failure rate in the ABC arm. Conversely, the ABC arm showed a lower rate of discontinuation of study medication owing to adverse events and a slightly more favourable lipid profile ^[6]. Here, we present the results of the resistance substudy based on genotype and phenotype testing in the subset of patients who experienced virological failure in one of the three study arms.

Methods

Patients

All HIV-positive patients recruited in the NEFA trial ^[6] who developed a virological failure (defined as two consecutive determinations of HIV RNA > 200 copies/ml separated by at least 2 weeks) were included in the substudy. In contrast to the main NEFA study, which published results at 1 year, the resistance substudy included data from patients with 2 years of follow-up. In order to determine the patterns of resistance mutations selected by treatment in any of the simplification arms (ABC, NVP or EFV), plasma samples were obtained at the detection of virological failure and stored at -70°C in each participating hospital and later sent to the resistance coordinator centre to perform genotypic and phenotypic resistance testing, as well as virtual phenotype.

Genotypic analysis

HIV-1 viral RNA was extracted from plasma, reverse transcribed into complementary DNA and subsequently amplified by a single tube polymerase chain reaction. HIV-1 mutations were detected using the ABI Viroseq HIV-1 Genotyping System (Applied Biosystems, Foster City, California, USA).

The genotypic results were interpreted using Retrogram™ (InferMed, London, UK) and revised by an expert advisory committee.

Phenotypic analysis

Plasma samples from all patients with virological failure were shipped to Virco, NV (Mechelen, Belgium). Phenotypic resistance was assessed using a recombinant virus assay by means of VIRCO Antivirogram ^[7]. Results were reported according to fold-change values, which reflect the fold increase in the mean 50% inhibitory concentration (IC₅₀) of a particular drug when it was tested with patient-derived recombinant virus isolates relative to the mean IC₅₀ of the same drug obtained when tested with a wild-type reference virus isolate.

Virtual phenotype analysis

Virtual phenotype was determined by Virco NV, according to protocols published elsewhere ^[8,9]. It used a mutation pattern-recognition method to retrieve the corresponding IC₅₀ values from a correlative database containing real phenotype data and the matching genotypic profiles.

Statistical analysis

Descriptive statistics have been used to report results. The χ² test was used to compare qualitative variables and κ concordance coefficient to compare genotype and phenotype results.

All nucleotide sequences were submitted to GeneBank and provided with accession numbers (AY833574 to AY833599).

Results

Of the 460 patients included in the study, 51 (11%) experienced virological failure after 24 months of follow-up while on assigned study medication. Of these, 12 out of 156 (8%) were in EFV group, 14 out of 155 (9%) were in NVP group and 25 out of 149 (17%) were in the ABC group (P = 0.04, χ² test). Virological failure was significantly more frequent in the subgroup of patients with previous history of monotherapy or dual therapy with NRTI [odds ratio (OR), 3.4; 95% confidence interval (CI), 1.7-7.2; P = 0.0002, by the χ² test] when compared with patients that initiated HAART-based therapy from the beginning (Table 1). Remarkably, subjects in the ABC arm with prior suboptimal therapy showed higher virological failure (29%) than subjects with prior suboptimal therapy in the NNRTI arms (13% for the NVP arm and 10% for the EFV arm). In contrast, there were similar levels of virological failures in the three study arms in subjects without prior suboptimal therapy (6%, 5% and 4% in ABC, NVP and EFV arms, respectively).

Table 1

The median time to virological failure was 36 weeks (range, 12-96): 38, 45 and 35 weeks for ABC, EFV and NVP arms, respectively.

Adequate plasma samples for resistance testing were available from 47 of the 51 patients (Table 2). HIV-1 RNA could be amplified in 30 of 47 samples (65%). Among the available samples in which HIV-1 RNA could not be amplified, 15 of 17 samples (88%) had a plasma viral load < 1000 copies/ml. In the phenotypic resistance analysis, recombinant virus yield was below the detection limit in 23 of 47 samples (51%). Consequently, genotype and phenotype resistance data were obtained in 23 patients and genotype data only in seven (Table 2).

Table 2

Genotypic resistance analysis

Among the 30 patients for whom HIV-1 RNA could be amplified, 18 (60%) were from the ABC group, seven (23%) from NVP group and five (16%) from EFV group (Table 2).

Two of these 30 patients (6%) failed therapy with a wild-type virus (one in the ABC and one in the EFV group). At least one thymidine analogue resistance mutation (M41L, D67N, K70R, L210W, T215F/Y, K219E/Q) was detected in 13/18 (72%) of the ABC group, 5/7 (71%) of the NVP group and 2/5 (40%) of the EFV group; the differences among the ABC and NNRTI arms were non-significant (OR, 1.86; 95% CI, 0.3-11.2; P = 0.46).

The K65R mutation was detected in only one patient, who was in the ABC group and also had the mutation S68G. Three patients had the L74V/I mutation, and one patient carried the mutations E44D and V118I.

The M184V mutation was detected in 24 out of 30 (80%) patients, 14 (58%) in ABC group, six (25%) in NVP group and four (16%) in EFV group. Again, differences among ABC and NNRTI arms were non-significant (OR, 0.70; 95% CI, 0.05-6.1; P = 1.0). Overall, the M184V/I mutation was detected in 24 of the 27 patients who had been treated with lamivudine. Two of the three patients failing while on lamivudine therapy and not carrying the M184V mutation had no detectable resistance mutations. The third patient without the M184V mutation was the one that carried the mutations E44D and V118I.

Of the 30 (87%) patients who had their HIV-1 RNA amplified, 26 out showed at least one polymorphism (between one and nine): protease region M36I, M46I, D60E, A62V, L63P, I64B, T69N, A71T, V77I, A98S; reverse transcriptase region K166R, H208I, R211K, L214F, P225H and K238T.

Mutations related to NNRTI (K101E/Q, K103N, V106A, V108I, Y181C, Y188L, G190A/S and G196E) were detected in 8 out of 12 patients failing while on treatment with either NVP or EFV. Among the remaining four patients, three carried M184V plus other nucleoside-resistance associated mutations (NAM), whereas the fourth patient showed no mutations in the genotypic resistance testing and had a high viral load (32 000 copies/ml). Finally 5/30 (16%) presented at least one resistance mutation in the gene for protease (L10I/V, M46L, I54V, V82A or D30N) at the time of virological failure. All had received suboptimal therapy (mono or bitherapy) prior to PI-containing HAART, had been randomized to receive simplification therapy with either NVP or EFV and also had some mutations conferring resistance to NNRTI.

Phenotypic resistance analysis

Both phenotypic and genotypic resistance tests were available for 23 patients.

There was good concordance between genotype and phenotype resistance test to lamivudine, NVP and EFV based on the κ concordance coefficient. In contrast, striking differences were found for stavudine, zidovudine, abacavir, didanosine and tenofovir, mostly because of a high proportion of samples considered as partially resistant in genotype interpretation and reported as sensitive in the real phenotype tests. The same results were found when genotype was compared with virtual phenotype (Table 3).

Table 3

Of the 23 patients with available phenotype, 14 were in the ABC group, five in the EFV group and four in the NVP group (Table 2). Five had received a PI-containing HAART as first-line therapy, and only two had a NAM. The remaining 18 patients had received suboptimal NRTI-containing therapy prior to PI-containing HAART (Tables 4 and 5). Four (17%) had a fold change in IC₅₀ above the normal susceptible range for ABC (cut-off for normal susceptible range, 3,0). One carried the K65R + S68G resistance mutation and another patient had three NAM plus M184V and L74V. The remaining three patients harboured M184V with two, four or six NAM, respectively (Table 4). Among the 18 patients with phenotype sensitive to ABC, nine patients carried one to five NAM plus M184V; four harboured only M184V; one carried two NAM; one carried five NAM; one carried E44D plus V118I plus four NAM; and two patients had no mutations (Tables 4 and 5). The only difference observed among the patients resistant or sensitive to ABC carrying M184V plus NAM was the presence or absence of mutation L210W. This difference was not found among those patients for whom a virtual phenotype was available.

Table 4

Table 5

There were no patients with a fold change in IC₅₀ to didanosine higher than the cut off of susceptibility (2.0). From a genotypic view, among those patients in whom real phenotype was available, two samples were considered resistant to didanosine because they carried L74V; the other 16 patients were considered to be partially resistant owing to the presence of M184V.

For stavudine (IC₅₀ cut-off for normal susceptible range, 1.8), only two patients had an IC₅₀ fold change to stavudine above the normal susceptible range by phenotype. One patient carried K65R and S68G and was considered sensitive by genotype as well as by virtual phenotype. The other patient had four NAM (M41L, D67N, L210W and T215Y) and also E44D and V118I, being considered resistant by genotype, real phenotype and virtual phenotype.

Discussion

One of the main results of the NEFA study was that there was a trend toward higher failure rates in the ABC group after switching therapy from a PI to NVP, EFV or ABC in virologically suppressed patients. Virological failures were almost exclusively concentrated among patients with prior suboptimal therapy with NRTI ^[6].

In this resistance substudy, cross-resistance between ABC and other NRTI drugs explained, at least in part, the higher rate of virological failure in the ABC group and also in the patients assigned to NVP or EFV who switched to ABC because of intolerance to NNRTI. The majority of patients (93%) with virological failure for whom HIV RNA could be amplified had genotypic resistance mutations. The expected proportion of resistance mutations among naive patients that failed on their first antiretroviral regimen is usually low. In the Combine substudy ^[10], 9 out of 16 (56%) naive patients who experienced virological failure had no evidence of genotypic resistance mutations. In contrast, Clumeck et al. ^[4] and Opravil et al. ^[3] reported that as few as 16% and 11%, respectively, of drug-experienced patients with virological failure after simplification had no resistance mutations identified, a proportion similar to that found in our study. It could be argued that the different results observed among both populations - naive and simplified treatment patients - could be caused by the fact that the proportion of non-adherent patients is much lower in a simplification trial than in naive patients; therefore, the majority of patients on simplified treatment will have resistance mutations when failure occurs.

There was a higher proportion of NAM among those patients experiencing virological failure in the ABC group than in the EFV or NVP groups. The majority of patients in the ABC group carried two or more NAM together with M184V. In contrast, among patients who failed in the NVP and EFV groups, there were fewer NAM detected but higher numbers of mutations conferring resistance to drugs with a low genetic barrier (Table 5); that is, drugs against which HIV-1 acquires high-level resistance with only a single mutation, such as lamivudine and NNRTI. As in the ABC group, the majority of patients who failed in the NVP and EFV groups had received prior suboptimal therapy.

Our results are in agreement with those reported in studies of induction therapy followed by maintenance therapy, where significantly fewer patients who continued to receive the induction regimens developed virological failure compared with those on a less-potent maintenance regimen ^[11-13]. In particular, treatment failures in the ACTG 343 study were seen mainly among patients previously treated with zidovudine, who harboured zidovudine-resistant HIV-1 ^[11].

Where genotypic and phenotypic analysis could be performed, resistance mutations were detected in almost all patients. These results could be explained by the fact that the majority of patients had been exposed to suboptimal antiretroviral therapy prior to PI-containing HAART. The higher rate of virological failure in patients randomized to ABC can be explained, at least in part, by cross-resistance between ABC and other NRTI, thus favouring the emergence of virus with preexisting resistance mutations to the replacement drugs.

Our results are in accordance with those of randomized trials of simplified maintenance therapy with ABC in patients treated with PI-based antiretroviral therapy ^[3,4]. Opravil et al. ^[3] have shown that a history of previous mono or dual zidovudine therapy could predict virological failure after the switch strategy. In most patients with virological failure, a wide range of mutations conferring resistance to reverse transcriptase were detected at the time of failure, but several of these mutations were already present at baseline when proviral DNA from peripheral blood mononuclear cells was analysed retrospectively.

The M184V lamivudine-related mutation was observed most frequently and, in fact, all but four patients with this mutation [26/30 (86%)] had been treated with a combination that included lamivudine. There were three patients treated with lamivudine but without the M184V mutation. Two of them had no mutation detected, indicating inadequate drug exposure owing to pharmacological factors or suboptimal patient adherence to drug therapy ^[14]. The remaining patient carried E44D and V118I, which have been associated with resistance to lamivudine when accompanied by several NAM in the absence of a concurrent M184V mutation ^[15,16]. This patient had also M41L, D67N, L210W and T215Y mutations.

One patient carried K65R and S68G, a combination frequently observed among those strains harbouring multidrug-resistance NRTI-related mutations ^[17,18]. This patient, randomized to the ABC arm, had been treated with didanosine plus stavudine, as in the study of Roge et al. ^[17]. This study evaluated the efficacy of the ABC plus stavudine plus didanosine combination; among the five patients without any resistance mutations at baseline, all carried K65R and four of them had also S68G when treatment failed.

The proportion of virological failure in the NVP and EFV groups was similar and lower than in the ABC group, as already shown in our main study ^[6]. NRTI- and NNRTI-related mutations showed similar patterns and were homogeneously distributed among both NVP and EFV arms; phenotypic analysis demonstrated also similar resistance patterns.

There was a good correlation between genotype and real phenotype resistance patterns. However, the results of real phenotype added valuable information (fold change in IC₅₀) mainly among those NRTI (ABC, stavudine, didanosine) showing possible resistance based on genotypic interpretation. In our study, the discrepancies between phenotype and genotype interpretation were mainly caused by the presence of the reverse transcriptase mutation M184V (n = 16). Its presence has been associated with a reduced virological response to didanosine and ABC ^[19], thus categorizing it as partially resistant from a genotypic point of view even though the samples were considered sensitive by phenotypic resistance assay. Recent reports have shown that the M184V contribution to didanosine resistance is quite weak, suggesting, as in our study, the inadequacy of considering partial resistance to didanosine if M184V is present, alone or together with other NAM ^[20,21]. With respect to ABC, L210W seems to play a discriminative role in phenotypic resistance assays, and its presence was related to a more than threefold increase in the IC₅₀. To our knowledge, only one study has shown that the L210W mutation is associated with a poor virological response to ABC ^[22]; however, in this study the investigators did not perform phenotypic resistance assays and there were no data about the remaining mutations present at the time of failure. The NRTI drugs zidovudine and stavudine share a similar resistance mutation pattern (thymidine analogue resistance mutations); however, there was a trend towards concordance of genotypic and phenotypic profile with zidovudine that can be explained by its more specific association with the T215Y resistance mutation. There seemed to be a good correlation of virtual phenotype with real phenotype.

Our study has an important limitation; no amplification could be performed for a substantial proportion of patients because of a low viral load (< 1000 copies/ml) at the time of virological failure. Analysis of proviral DNA would have been of interest in this particular setting in order to obtain additional data to test our hypothesis that the majority of patients who had virological failure had preexisting NRTI-related resistance mutations.

In summary, cross-resistance mutations involving NRTI can explain the higher risk of virological failure in patients switched to ABC compared with switching to NVP- or EFV-containing HAART, particularly if the patient had received suboptimal antiretroviral therapy prior to PI-containing HAART. Phenotypic resistance testing may be helpful to confirm those genotypic results that are difficult to interpret. However, further considerations are required to improve the interpretation of the role of M184V in some NRTI genotypic resistance results.

Sponsorship: This study was supported by the Comisión Interministerial de Ciencia y Tecnología (SAF 1998-0021 and SAF 2001-2591), Fondo de Investigación Sanitaria, Ministerio de Sanidad y Consumo (PIO 2590 and Red Temática Cooperativa de Investigación en SIDA RIS G03-173.