| |
The Impact of Transmitted Drug-Resistance on Treatment Selection and Outcome of First-Line Highly Active Antiretroviral Therapy (HAART)
|
| |
| |
JAIDS Journal of Acquired Immune Deficiency Syndromes:
April 2010 - Volume 53 - Issue 5 - pp 633-639
Bansi, Loveleen MSc; Geretti, Anna Maria MD; Dunn, David PhD; Hill, Teresa PhD; Green, Hannah MSc; Fearnhill, Esther; Gazzard, Brian Prof; Nelson, Mark MD; Porter, Kholoud PhD; Phillips, Andrew Prof; Sabin, Caroline Prof; and on behalf of the UK Collaborative Group on HIV Drug Resistance UK Collaborative HIV Cohort (CHIC) Study
From the *University College Medical School, London, United Kingdom; Royal Free Hampstead NHS Trust, London, United Kingdom; Medical Research Council Clinical Trials Unit, London, United Kingdom; and Chelsea and Westminster Hospital, London, United Kingdom.
Abstract
Objective: The study aim was to determine how resistance testing influences outcome of first-line highly active antiretroviral therapy (HAART) in routine practice in the United Kingdom.
Methods: The prevalence of transmitted drug resistance and the genotypic sensitivity score (GSS) of first-line HAART regimens were determined using data from the UK Collaborative HIV Cohort (CHIC) Study. Factors associated with starting a regimen with a reduced GSS and subsequent virological responses were analyzed by logistic and Cox regression.
Results: Amongst patients tested in 1999-2006, 116 of 1175 (10%) had ≥1 resistance mutation; 64 patients (5.4%) had ≥1 mutation associated with resistance to drugs in the initial HAART regimen and 54 (4.6%) showed a GSS <3. Factors independently associated with a GSS <3 were starting HAART in 1999-2001 vs. 2004-2006 (odds ratio = 2.63; 95% confidence interval: 1.19 to 5.83) and use of ritonavir-boosted protease inhibitor (PI/r)-based vs. nonnucleoside reverse transcriptase inhibitor-based regimens (1.97; 1.06 to 3.64). A GSS >3 was independently associated with virological suppression (hazard ratio for GSS <3 = 0.60; 95% confidence interval 0.41 to 0.87).
Conclusions: Most patients starting HAART after undergoing resistance testing received regimens with a GSS ≥3. PI/r-based therapy was often selected in patients with resistance to the nucleoside reverse transcriptase inhibitor backbone. Low GSS predicted poor virological suppression and the association persisted after adjusting for PI/r use.
INTRODUCTION
In Western Europe and North America, 10%-15% of HIV-infected patients have evidence of transmitted drug resistance (TDR), most commonly to the nucleos(t)ide and nonnucleoside reverse transcriptase inhibitors (NRTI and NNRTIs).1-8 There is growing concern that TDR may compromise responses to highly active antiretroviral therapy (HAART), especially in the context of NNRTI-based regimens.5,9-12 The British HIV Association and the International AIDS Society recommend that all newly diagnosed HIV-infected individuals undergo resistance testing to determine the presence of TDR and inform the selection of first-line HAART.13,14 However, drug resistance is only one of a number of factors that may be considered when choosing an initial regimen, as concerns about possible toxicities and likely adherence, and patients' views on the ability of any particular regimen to accommodate lifestyle requirements, are major determinants of the selection of first-line HAART. In some cases, information about TDR may only become available after the patient has started treatment. In other cases, the significance of certain genotypic changes may be missed, if phenotypic tests are used.15 The overall resilience of the regimen to rapidly accumulating drug resistance-referred to as the "genetic barrier"-may also lead to a preference for regimens based on ritonavir-boosted protease inhibitors (PI/r) rather than NNRTIs as a compensation for the use of a partially active NRTI backbone. These considerations may lead to some patients starting regimens that, according to their baseline resistance test, contain fewer than 3 fully active drugs. The extent to which virological responses are influenced by these choices is unclear.
The aim of this study was to investigate TDR in a large cohort who underwent genotypic resistance testing whilst HAART-naive and subsequently started first-line therapy. We determined the genotypic sensitivity score (GSS) of the initial regimen and assessed the impact of GSS on virological outcomes, including the initial plasma viral load (VL) suppression <50 copies per milliliter and the subsequent risk of virological rebound.
DISCUSSION
We investigated the use of resistance testing among drug-naive patients in routine care in the United Kingdom and showed that testing was biased in favor of white homosexual males during 1999-2006. This reflects the higher risk of resistance traditionally observed in this group relative to heterosexuals, women, and persons of black ethnicity.18 It is only in recent years that national guidelines have started to recommend testing in all newly diagnosed patients regardless of risk group and ethnicity.19 The prevalence of TDR was 9.9%, consistent with previous estimates from the United Kingdom and elsewhere in Western Europe.5,7,18,20-22 TAMs for the NRTIs and K103N for the NNRTIs were the most prevalent mutations, in line with earlier findings.1 Importantly, most patients started a regimen comprising at least 3 fully active drugs. Although only a small minority (4.6%) started regimens with a GSS <3, this group experienced reduced virological responses and were less likely to achieve an undetectable VL than patients starting fully active regimens. There was no significant effect of GSS on the risk of virological rebound after suppression, possibly reflecting the small number of patients who experienced rebound.
The proportions of patients who started a PI-based or NNRTI-based regimen despite having mutations conferring resistance to these drugs were in line with those reported by Yerly et al in 200721 and Masquelier in 20056 at around 2% and 2.5%, respectively. The proportion of patients with mutations conferring resistance to the NRTIs was also similar at 4% compared with 6% in the study by Yerly et al.21 We found that persons with TAMs were less likely to receive zidovudine, reflecting the well-recognized resistance effect of these mutations on the drug. However, there was no negative impact of having TAMs on the likelihood of receiving other NRTIs that can also be affected by these mutations, including stavudine, abacavir, didanosine, and tenofovir.15 The finding that a small proportion of patients started NNRTI-based regimens despite the presence of NAMs or nNAMs is surprising based on the current understanding of the low genetic barrier of such regimens.19 Given that starting HAART in earlier calendar years was independently associated with a GSS <3, one explanation for these findings may be that in earlier years the impact of TDR was underappreciated. In addition, resistance interpretation algorithms have evolved over time, together with an expanding range of drug options. Thus, it may be proposed that the availability of TDF in later calendar years may have contributed to the higher GSS seen in this period.
Interestingly, starting HAART with PI/r-based regimens was significantly associated with a GSS <3. The reduced GSS in these patients was infrequently related to the presence of PRAMs. Rather, most patients had NAMs affecting the NRTI backbone. These findings suggest that in a cohort setting where NNRTI-based therapy is the generally preferred initial HAART regimen,23 PI/r were selected specifically to compensate for the presence of NRTI resistance. A further consideration is that we used an interpretation algorithm, which assigned a GSS of 1 to a fully active PI/r. It has been proposed that a fully active PI/r should be given a score of 1.5.24 Using a higher GSS score for PI/r increased the overall GSS for each patient with TDR, but only minimally changed the number of patients in the GSS <3 category (from 54 to 44 overall).
Importantly, however, a GSS <3 was found to be an independent predictor of reduced virological suppression, even after adjusting for starting HAART with a PI/r. This suggests that starting HAART with PI/r (and in particular, lopinavir/r in this cohort) did not fully compensate for the presence of resistance to the NRTI backbone. Although this finding requires further confirmation, it is consistent with the superior virological activity of lopinavir/r in combination with 2 NRTIs relative to lopinavir/r monotherapy in first-line HAART.25
There are limitations to our study. Not all patients underwent resistance testing before starting HAART and selection in favor of white homosexual males was evident. We analyzed an observational cohort and cannot exclude an effect of unmeasured confounders that may contribute to both a low GSS and a poor virological outcome on first line. Further, we are unable to fully investigate the viral evolution of patients with GSS <3 who experienced virological failure.
In summary, we have shown that most but not all patients starting HAART do so on fully active regimens. This number is likely to increase with the availability of new drugs in recent calendar years. However, we have shown that despite an attempt at compensating for the presence of NRTI TDR by using a PI/r-based regimen, patients with a GSS <3 showed poorer virological outcomes than those starting on fully active regimens. These findings indicate that selection of first-line HAART should take into account the presence of TDR, together with other recognized predictors of virological outcomes. The contribution of NRTI TDR to virological failure of PI/r-based regimens has potential clinical significance and warrants further investigation.
RESULTS
Resistance Testing in Routine Practice
Among 10,668 patients who started first-line HAART in 1999-2006, 1175 (11.0%) underwent pre-HAART resistance testing as part of routine care (Table 1). Compared with these patients, those who did not undergo testing before starting HAART were more likely to be heterosexual (94% vs. 85% in homosexual, P < 0.0001), female (93% vs. 87% of males, P < 0.0001), and black African (94% vs. 86% of white people, P < 0.0001). They also had lower median pre-HAART CD4 counts (180 vs. 218 cells/mm3, P < 0.0001), though there was no difference in the median pre-HAART VL (5.0 log10 copies/mL, P = 0.13).
Prevalence of TDR
Overall, 116 of 1175 patients (9.9%) had ≥1 mutation indicative of TDR, with 43 of 116 (37.1%) showing >1 mutation (Table 2). The most common mutations were the TAMs T215 revertants (n = 41, 3.5%) and M41L (n = 24, 2.0%), the nNAM K103N (n = 29, 2.5%), and the PRAM L90M (n = 12, 1.0%). The median (range) number of mutations amongst those with ≥1 mutation was 1 (1-9); patients had a median 1 (0-6) NAMs, 0 (0-3) nNAMs, and 0 (0-4) PRAMs.
Among the 1175 starting HAART, 64 (5.4%) had ≥1 mutation associated with resistance to the drugs used in their regimen (Table 3). This included 51 patients (4.3%) starting NRTIs, 3 of 111 (2.7%), 13 of 749 (1.7%) patients starting nevirapine or efavirenz, respectively, and 7 of 273 patients (2.6%) starting a PI (boosted or nonboosted). Five patients had resistance mutations for both NRTIs and PIs in their regimen, and a further 5 patients had resistance mutations for both NRTIs and NNRTIs in their regimen.
GSS of first-Line Regimen
The GSS of the first-line regimen was 3 in 1066 of 1175 patients (90.7%), >3 in 55 of 1175 patients (4.7%), and <3 in 54 of 1175 patients (4.6%), the latter comprising scores of 1, 1.5, 2, and 2.5 in 5 (0.4%), 1 (0.1%), 17 (1.4%), and 31 (2.6%) patients, respectively. Amongst the 1112 patients starting HAART with 3 drugs, scores were 3, 2.5, 2, 1.5, and 1 in 1065 (95.8%), 27 (2.4%), 15 (1.3%), 1 (0.1%), and 4 (0.4%) patients, respectively. Amongst the 63 patients who started HAART with >3 drugs, scores were >3, 2.5, 2, and 1 in 57 (90.5%), 4 (6.3%), 1 (1.6%), and 1 (1.6%) patients, respectively.
The 54 of 1175 patients (4.6%) who started a regimen with a GSS <3 had similar demographic characteristics and CD4 counts as patients with a GSS ≥3 (Table 4). However, patients with a GSS <3 tended to have had their resistance test performed in earlier calendar years, had a lower VL both at the time of test and at HAART initiation, and were more likely to start a PI-based rather than an NNRTI-based regimen.
In unadjusted analyses, factors significantly associated with a GSS <3 were starting HAART in earlier calendar years [odds ratio (OR) 2.50; 95% confidence interval (CI): 1.15 to 5.42 for 1999-2001 vs. 2004-2006] and starting PI/r (OR: 1.86, 95% CI: 1.02 to 3.42) compared with NNRTI-based HAART. In adjusted analyses, both variables remained significantly associated with a GSS <3, with adjusted ORs of 2.63 (95% CI: 1.19 to 5.83) and 1.97 (95% CI: 1.06 to 3.64), respectively.
Given that patients starting HAART with PI/r were more likely to have a GSS <3, these patients were investigated in greater detail. Of the 54 patients showing a GSS <3, 17 started PI/r-based HAART. Analyzing the resistance mutations of these 17 patients, 5 had mutations to the PI, whereas the other 12 patients resistance to the NRTIs in their initial regimen, most commonly to 1 NRTI (5 patients to tenofovir, 2 to zidovudine, 2 to didanosine, and 1 to abacavir) and in 2 patients to both NRTIs (tenofovir/abacavir and tenofovir/zidovudine, respectively). All 5 patients with PI resistance started lopinavir/r (2 with low-level and 3 with intermediate resistance to the drug); 1 patient with intermediate resistance to lopinavir/r also had resistance to, and started saquinavir.
In both unadjusted and adjusted models, patients were less likely to start an NNRTI if they had nNAMs (adjusted OR = 0.20; 95% CI: 0.10 to 0.39), whereas there was no evidence that patients with PRAMs were less likely to start a PI (adjusted OR = 2.02; 95% CI: 0.83 to 4.94). In the presence of TAMs, patients were less likely to start zidovudine (adjusted OR = 0.27; 95% CI: 0.12 to 0.64) and more likely to start didanosine (adjusted OR = 2.63; 95% CI: 1.14 to 6.06), but there was no significant relationship with starting the other TAM-affected drugs, stavudine, abacavir, and tenofovir.
Virological Suppression
Among 935 patients for whom both baseline and follow-up VL measurements were available, 763 patients (81.6%) achieved a VL <50 copies per milliliter after median 3.5 (interquartile range: 2.4-5.5) months (Fig. 1). The majority of these patients [573 of 763 (75.1%)] were receiving NNRTI-based regimens; 154 of 763 (20.2%) were receiving PI-based regimens; and 36 of 763 (4.7%) were receiving triple NRTI regimens. Among the 763 patients who achieved virological suppression, the GSS was ≥3, 2.5, 2, and 1 in 732 (95.9%), 19 (2.5%), 10 (1.3%), and 2 (0.3%) patients, respectively. Of the 172 patients (18.4%) who did not achieve virological suppression, the GSS was ≥3, 2.5, 2, 1.5, and 1 in 163 (94.8%), 4 (2.3%), 3 (1.7%), 1 (0.6%), and 1 (0.6%) patients, respectively.
Amongst the 40 patients with GSS<3, 31 (77.5%) achieved virological suppression, whereas 9 (22.5%) did not achieve virological suppression (P = 0.49). The HAART regimens of the 31 patients who achieved virological suppression despite a GSS <3 were NNRTI-based in 18 patients (9 with NAMs, 7 with nNAMs, and 2 with NAMs, and nNAMs), PI-based in 1 patient with NAMs, PI/r-based in 9 patients (6 with PRAMs and 3 with NAMs and PRAMs), and triple NRTIs in 3 patients with NAMs. Of the 9 patients with GSS <3 who did not achieve virological suppression, 6 of 9 started NNRTI-based regimens (1 with NAMs, 3 with nNAMs, and 2 with NAMs and nNAMs). The remaining 3 of 9 patients started PI-based regimens (lopinavir/r, nelfinavir, and atazanavir/r); none had PRAMs although all 3 had NAMs. Four of the 5 patients with resistance to the PI in their initial regimen were included in this analysis. They started HAART with lopinavir/r, and all achieved virological suppression, including 2 patients who changed from lopinavir/r to atazanavir/r and saquinavir/r, respectively, before reaching virological suppression.
Several Cox models were fitted to identify factors independently associated with achieving an undetectable VL after starting HAART. In a multivariable model in which calendar year is considered a confounder (Table 5), patients with a GSS <3 were less likely to achieve virological suppression compared with patients with a GSS ≥3 (P = 0.001). Patients starting HAART in earlier calendar years, those with higher pre-HAART VL and those starting nonboosted PI or triple NRTI regimens compared with an NNRTI-based regimen were less likely to achieve virological suppression. A marginally significant relationship was also seen between lower pre-HAART CD4 counts and achieving an undetectable VL. The GSS was not significantly associated with virological suppression if analyses were restricted to patients with GSS of ≥3.
An interaction test showed a significant association between calendar year and GSS (P = 0.01). Hence, sensitivity analyses were performed in which models were stratified by the year in which the resistance test was performed. A higher GSS was significantly associated with achieving virological suppression when resistance tests were performed in both earlier and later calendar years: the hazard ratio per 1 unit increase was 1.96 (95% CI: 1.11 to 3.48) for 1999-2001 and 2.66 (95% CI: 1.23 to 5.73) for 2004-2006; there was also an association with tests performed in 2002-2003 although the confidence interval was large (hazard ratio = 1.26; 95% CI: 0.94 to 1.70, per 1 unit increase).
Virological Rebound After Achieving Suppression
Among 722 patients who achieved virological suppression within a year of starting HAART, virological rebound occurred in 68 (9.4%), at median 5.0 months after suppression. This group included 63 patients with 2 consecutive VL measurements >400 copies per milliliter and 5 with a single VL >400 copies per milliliter followed by a treatment change. Of patients with a GSS<3, 31 were included in these analyses, 6 of which experienced virological rebound at median 3.0 months. All 6 had started HAART on NNRTI-based regimens and 4 of 6 had nNAMs, whereas the other 2 patients had resistance to the NRTIs in their initial regimen.
GSS was not associated with virological rebound (OR: 0.97; 95% CI: 0.48 to 1.97, P = 0.93). In adjusted analyses, factors associated with virological rebound included younger age (adjusted OR: 0.71; 95% CI: 0.51 to 1.01, per 10 years older), higher pre-HAART CD4 counts (1.11; 95% CI: 1.02 to 1.20, per 50 cells/mm3 higher), switching therapy for any reason before failure (adjusted OR: 4.86; 95% CI: 2.71 to 8.72), and starting a nonboosted PI-based regimen (adjusted OR: 2.52; 95% CI: 0.98 to 6.45) compared with an NNRTI-based regimen.
METHODS
Study Population
Genotypic resistance test results and linked clinical data from 10 centres were derived from the UK HIV Drug Resistance Database and the UK Collaborative HIV Cohort (CHIC) Study. The UK HIV Drug Resistance Database was established in 2001 as a central repository of resistance tests performed as part of routine clinical care in the United Kingdom. The UK CHIC Study is an observational cohort of HIV-infected individuals attending some of the largest HIV clinical centres in the United Kingdom (Cited Here...).16
Eligible patients started first-line HAART (defined as a standard regimen of ≥3 drugs including NRTIs in combination with either 1 NNRTI, 1 or 2 PIs or another NRTI) between January 1999 and April 2006. Patients receiving combinations without NRTIs or with both protease inhibitors (PIs) and NNRTIs were excluded. Those who had a VL <50 copies per milliliter before starting HAART were also excluded as an undetectable pre-HAART VL may indicate either missing or inaccurate records.
Amongst patients who had a resistance test performed, mutations indicative of TDR were identified according to the 2007 Shafer "Resistance surveillance list,"17 whereas all mutations detected were used to predict drug susceptibility by the Stanford interpretation algorithm (July 2007). The proportion of patients with any mutation and with mutations to any drug used in their initial regimen was calculated, with separate analyses for resistance-associated mutations for the NRTIs (NAMs), the thymidine analogues (TAMs), the NNRTIs (nNAMs), and the PIs (PRAMs). GSS were determined by assigning a score of 1 to susceptible/potential low-level resistance, 0.5 to low-level/intermediate resistance, and 0 to high-level resistance.
Statistical Analysis
The characteristics of the HAART-naive patients who underwent resistance testing performed in the study period were compared with those who did not undergo testing using χ2 and Mann-Whitney tests. Logistic regression was used to identify factors associated with starting a first-line regimen with a GSS <3, adjusting for the following potential confounders: age; gender; ethnicity; HIV transmission risk group; HIV-1 subtype (B or non-B); calendar year of HAART initiation; interval between the resistance test and start of HAART, pre-HAART, HAART, CD4, and VL; and HAART regimen [PI-based, NNRTI-based or 'other' (triple NRTIs)]. PI-based therapy included both nonboosted and ritonavir-boosted PIs. Logistic regression was also used to investigate whether patients with nNAMs, PRAMs, or TAMs were less likely to start an NNRTI-based regimen, a PI-based regimen, or the TAM-affected NRTIs zidovudine, stavudine, abacavir, didanosine, and tenofovir, respectively.
Amongst patients who had a pre-HAART VL and at least 1 follow-up VL, Cox regression was used to identify factors associated with the rate of achieving a VL <50 copies per milliliter within a year of starting HAART. Rates of virological rebound after suppression (defined as 2 consecutive VL measurements >400 copies per milliliter or 1 VL >400 copies per milliliter followed by any treatment change) were also calculated using Cox regression, where only patients with at least 1 follow-up VL measurement after suppression were included. In these models, GSS was fitted as a continuous variable because a linear trend between GSS and the outcome of interest was observed in univariate analyses. In addition, a time-updated covariate incorporating any switch in therapy before end of follow-up was included in both the virological suppression and rebound analyses.
| |
| |
| |
|
|
|
