Abstract
Objectives:
The main objective of this analysis was to evaluate the impact of pre-existing drug resistance ( PDR) by next generation sequencing (NGS) on the risk of treatment failure (TF) of first-line regimens in participants enrolled in the START study.
Methods:
Stored plasma from participants with entry HIV RNA>1,000 copies/ml were analysed by NGS (llumina MiSeq). PDR was defined using the mutations considered by the Stanford HIVDB (v8.6) to calculate the genotypic susceptibility score (GSS, estimating the number of active drugs) for the first-line regimen at the detection threshold windows of >20%, >5% and >2% of the viral population. Survival analysis was conducted to evaluate the association between the GSS and risk of TF (VL>200 copies/mL plus treatment change).
Results:
Baseline NGS data was available for 1,380 ART-naïve participants enrolled over 2009–2013. First-line ART included an NNRTI in 976 (71%), a PI/r in 297 (22%), or an INSTI in 107 (8%). The proportion of participants with GSS<3 were: 7% for >20%, 10% for >5% and 17% for the >2% thresholds, respectively. The adjusted hazard ratio (HR) of TF associated with a GSS of 0–2.75 vs. 3 in the subset of participants with mutations detected at the >2% threshold was 1.66 (95% CI:1.01–2.74, p=0.05) and 2.32 (95% CI:1.32–4.09, p=0.003) after restricting the analysis to participants who started an NNRTI-based regimen.
Conclusions:
Up to 17% of participants initiated ART with a GSS<3 on the basis of NGS data. Minority variants were predictive of TF, especially for participants starting NNRTI-based regimens.
Keywords: Human Immunodeficiency Virus (HIV), HIV drug resistance, Antiretroviral therapy, Next generation sequencing
INTRODUCTION
Human immunodeficiency virus type 1 (HIV-1) drug resistance genotyping is recommended to help guide selection of antiretroviral therapy (ART) and to prevent virologic failure. Although DRMs are typically detected by Sanger sequencing if present in 15–25% of the total viral population [1–3], more sensitive techniques have been developed [4–8] with an increased sensitivity range.
Minority variants have been shown to be associated with the risk of failing ART, but the strength of the associations is variable across drug classes (strongest for nonnucleoside reverse transcriptase inhibitors (NNRTIs) and CCR5 antagonists). In contrast, for nucleoside reverse transcriptase inhibitors (NRTIs) and the integrase strand transfer inhibitor (INSTI) raltegravir the evidence is only moderate and very low in the case of the protease inhibitors (PIs), as well for the INSTI elvitegravir, and dolutegravir [9–13].
A pooled analysis of data from 10 studies showed that people with minority variants detected at baseline had 2-fold higher risk of virological failure compared to participants without these variants [14]. Of note, the increased risk of virological failure appeared to be mainly driven by NNRTI-resistant minority variants and was independent of medication adherence. These results were confirmed in several subsequent studies conducted in other settings [15–21]. However, a more recent qualitative review identified a total of 25 studies that looked at the association between detection of minority variants and risk of virological failure of a NNRTI-based first line regimen with conflicting findings [22].
In the light of these newer findings, the aim of this analysis was to re-evaluate the impact of minority variants using next generation sequencing (NGS), including the role of NNRTI associated resistance, on the risk of failure of first-line regimens initiated by the target population enrolled in the START trial.
METHODS
Study population and sequencing
The START trial, conducted by the International Network for Strategic Initiatives in Global HIV Trials (INSIGHT), enrolled ART-naïve HIV-positive participants between April 2009 and December 2013. The study design and data collection plan for START has previously been reported [23,24]. A plasma sample, taken within 60 days prior to enrolment was obtained from all participants who provided consent for stored specimens.
Methods for samples preparation, amplification of viral RNA and NGS, identification of drug resistance mutation by means of VirVarSeq and determination of HIV-subtype have been described elsewhere [25,26].
Sequence reads (FASTQ files) were analysed with VirVarSeq version 20140929, which calls variants at the codon level [27]. From the output, we extracted amino acid frequencies in pol gene from amino acid position 1 to 935 where positions 1–99 encode protease (PR) protein, positions 100–659 encode reverse transcriptase (RT) protein, and positions 660–935 partially encode integrase (IN) protein (our amplicon did not cover position 936–947) [25].
Definition of pre-existing drug resistance and phenotypic drug susceptibility
In this analysis, we considered all mutations used by the Stanford HIVdb algorithm v8.6 to provide genotypic test results interpretation. Of note, the Stanford HIVdb algorithm considers a much wider range of mutations than those considered by the WHO 2009 surveillance list used in previous analyses of these dataset [27–29]. The Stanford HIVdb interpretation was used to calculate a genotypic susceptibility score (GSS) for the drugs included in participants’ first line regimen as follows [30] : each drug included in the initial regimen was given an individual numerical score of 1 if the interpretation was ‘no resistance’ (fully active drug), 0.75 if potential low level, 0.5 if low level, 0.25 if intermediate and 0 if high level of resistance (zero activity of the drug). The GSS for a regimen was the sum of the individual scores for the drugs included in the regimen and estimated the total number of drugs predicted to be active. For a triple combination regimen GSS varies between 0 and 3. Detail regarding our choice for sequencing depth and thresholds for calling DRMs are reported in the Appendix.
Statistical Analysis
Characteristics of participants were described and compared across GSS strata, grouped as 0–2.75 vs. 3 (i.e. partially vs. fully active regimen ) at the >20% threshold. The distribution of categorical variables was compared using a chi-square test and continuous variables using the U Mann-Whitney test. The breakdown of the exact first-line treatment was also shown overall and by GSS strata. We also described the distribution of the HIVDB interpretation scores separately by antiretroviral drug and according to threshold window used.
The primary endpoint of this analysis was treatment failure (TF), a composite outcome defined at the time of two possible failure events: 1) the time of experiencing a single plasma HIV-RNA>200 copies/mL after >6 months of therapy initiation if this event was followed by a change of ≥1 drugs (including discontinuation) of the regimen received at time of the elevated HIV-RNA value or 2) the time of pure confirmed virological failure (VF), defined at the time of the first of two consecutive HIV-RNA viral loads >200 copies/mL. For both components of the primary endpoint, survival time accrued from the date of therapy initiation until the date of the failure event or the date of the last available HIV-RNA measure, whichever occurred first. All available viral loads were used and the main assumption behind the definition 1) is that therapy changes occurring after having observed a viral load>200 copies/mL were with the intent of re-suppressing the viral load.
Standard unweighted Kaplan-Meier and Cox regression model analysis were performed. Suboptimal GSS was defined as having less than 3 active drugs. The association between GSS (fitted as a binary covariate 0–2.75 vs. 3 at the various thresholds) and the risk of experiencing the endpoints was evaluated in univariable analyses and after controlling for a number of variables chosen from the following initial list: age, gender, geographical region, mode of HIV transmission, hepatitis co-infection status, baseline CD4 count and HIV-RNA, trial arm, non-adherence, HIV subtype, year of starting ART and type of ART started. We used a direct acyclic graph (DAG) to describe our assumptions regarding the underlying causal structure of the data. On the basis of these assumptions, only baseline HIV-RNA, geographical region, intervention arm of START (immediate vs. deferred ART) and year of starting ART were deemed as common causes of both the exposure of interest and outcome, so were included in the multivariable models as potential confounders (Supplementary Figure S1).
A first sensitivity analysis was performed after restricting to participants with Sanger GSS≥2 (the subset in which NGS is likely to provide additional benefit, n=1,349, Table S1B). A second sensitivity analysis was performed using an alternative endpoint of treatment failure, which, for the component 1), only counted the discontinuations of ART following failure as events. Reasons for discontinuing drugs as coded by the treating physicians were summarized. Linear regression analysis controlling for extra-sample error distribution by means of cross-validation was also used to identify mutations at the 2–20% level associated with the week 4 decline in HIV-RNA. Hazard ratios (HR) of TF associated with these identified mutations were also calculated. More details for the linear regression analysis are shown in Supplementary material. A two-sided test of less than 0.05 was considered statistically significant. All statistical analyses were performed using the SAS software, version 9.4 (Carey NC, USA).
RESULTS
Overall, out of the 1,819 with NGS data at the ≥200 read depth, 1,380 (76%) who started ART with ≥3 drugs either in the immediate or deferred arm and had virological follow-up of ≥1 month were included in this analysis. Table 1 shows the main characteristics of the study population, a selection of factors potentially associated with the risk of virological failure to HIV treatment, stratified by GSS groups using the 20% detection threshold. Of note, 96/1,380 participants (7%) started a regimen predicted by GSS to include <3 active drugs at this threshold. The two GSS groups appeared to be balanced with respect of demographics and immune-virological factors. The only marked difference were in region of enrollment with South America being over-represented in the group with a GSS of 0–2.75. In contrast, participants in Africa and the USA appeared to be more likely to have a GSS=3 (chi square p<.001). In addition, the median time from HIV diagnosis for participants in the GSS of 0–2.75 group was on average 3 months shorter than that of the GSS=3 group (p=0.02, Table 1). Of note, participants’ characteristics were similar when other window thresholds were used to define the GSS groups (see Table S1A for the >2% threshold for example). The proportion of participants with a GSS of 0–2.75 using this threshold was increased at 17.0% ( 235/1380).
Table 1.
Characteristics of participants according to HIVDB GSS (>20% threshold)
| Characteristics | 0<GSS#<=2.75 | GSS#=3 | p-value$ | Total |
|---|---|---|---|---|
| N= 96 | N= 1284 | N= 1380 | ||
| Gender | 0.992 | |||
| Female | 15 (15.6%) | 201 (15.7%) | 216 (15.7%) | |
| Race | 0.258 | |||
| Asian | 7 (7.3%) | 90 (7.0%) | 97 (7.0%) | |
| Black | 10 (10.4%) | 246 (19.2%) | 256 (18.6%) | |
| White | 57 (59.4%) | 724 (56.4%) | 781 (56.6%) | |
| Hispanic | 18 (18.8%) | 199 (15.5%) | 217 (15.7%) | |
| Other | 4 (4.2%) | 36 (2.8%) | 40 (2.9%) | |
| Mode of HIV transmission | 0.553 | |||
| Heterosexual contacts | 19 (19.8%) | 327 (25.5%) | 346 (25.1%) | |
| PWID | 1 (1.0%) | 19 (1.5%) | 20 (1.4%) | |
| MSM | 73 (76.0%) | 886 (69.0%) | 959 (69.5%) | |
| Other/unknown | 3 (3.1%) | 52 (4.0%) | 55 (4.0%) | |
| Region of enrollment | <.001 | |||
| Africa | 3 (3.1%) | 136 (10.6%) | 139 (10.1%) | |
| Asia | 7 (7.3%) | 78 (6.1%) | 85 (6.2%) | |
| Australia | 5 (5.2%) | 40 (3.1%) | 45 (3.3%) | |
| Europe | 29 (30.2%) | 567 (44.2%) | 596 (43.2%) | |
| South America | 46 (47.9%) | 309 (24.1%) | 355 (25.7%) | |
| USA | 6 (6.3%) | 154 (12.0%) | 160 (11.6%) | |
| Age, years | 0.884 | |||
| Median (IQR) | 35 (28, 47) | 35 (28, 43) | 35 (28, 44) | |
| CD4 count, cells/mm3 | 0.177 | |||
| Median (IQR) | 654 (588, 730) | 632 (575, 721) | 633 (575, 721) | |
| HIV-RNA, log10 copies/mL | 0.630 | |||
| Median (IQR) | 4.45 (4.00, 4.79) | 4.48 (4.01, 4.86) | 4.47 (4.01, 4.86) | |
| CD4/CD8 ratio | 0.208 | |||
| Median (IQR) | 0.6 (0.5, 0.9) | 0.6 (0.4, 0.8) | 0.6 (0.4, 0.8) | |
| HBsAg, n(%) | 2 (2.1%) | 34 (2.7%) | 0.720 | 36 (2.7%) |
| HCVAb, n(%) | 1 (1.1%) | 36 (2.9%) | 0.299 | 37 (2.7%) |
| BMI, Kg/m2 | 0.645 | |||
| Median (IQR) | 23.8 (22.0, 26.6) | 24 (22, 27) | 24 (22, 27) | |
| Calendar year of starting ART | ||||
| Median (IQR) | 2012 (2011,2013) | 2012 (2011,2013) | 0.55 | 2012 (2011,2013) |
| Months from HIV diagnosis to enrolment | 0.015 | |||
| Median (IQR) | 8 (3, 22) | 11 (4, 31) | 10 (4, 30) | |
| Months from GRT* test to start of ART | 0.286 | |||
| Median (IQR) | 12 (2, 28) | 12 (3, 30) | 12 (3, 29) | |
| Co-morbidities, n(%) | ||||
| Cardiovascular disease | 2 (2.1%) | 8 (0.6%) | 0.104 | 10 (0.7%) |
| Diabetes | 2 (2.1%) | 37 (2.9%) | 0.649 | 39 (2.8%) |
| Dyslipidemia | 2 (2.1%) | 53 (4.1%) | 0.323 | 55 (4.0%) |
| Hypertension | 7 (7.3%) | 125 (9.7%) | 0.433 | 132 (9.6%) |
Chi-square or Mann-Whitney U as appropriate
HIVDB v8.6 with >20% threshold
Genotypic Resistance Test
Supplementary Figure S2 shows the breakdown of antiretroviral regimens with the large majority of the population initiating a triple drug regimen which included TDF/FTC (n=80, 83% of those with GSS=0–2.75 and 1,148, 89% of those with GSS=3) in combination with efavirenz (77% of the GSS 0–2.75 group vs. 63% of the GSS=3 group).
Figure 1 (panels A, B, and C) shows the breakdown of the Stanford predictions for drug components of the initial regimen of participants, stratified by the threshold window used to define resistance. Using the 20% Sanger threshold, participants’ viruses showed high-level resistance to efavirenz (3%) and 3TC or FTC (0.7 and 0.2%, respectively) while there appeared to be intermediate resistance to elvitegravir (6%) but not to the other INSTI. Interestingly, as seen in previous analyses of this same dataset, by lowering the threshold for resistance detection to >2%, a higher percentage of participants appeared to have a virus with high level resistance to raltegravir (1.5%) and 1–3% retaining high level resistance to 3TC/FTC and 6% to RPV and efavirenz. Because a large proportion of mutations were detected at very low levels, more resistance (at any level) was detected at the >2% threshold (Figure 1C ) compared to the >5% threshold (Figure 1B ).
Figure 1. HIVDB interpreted drug activity <1 according to drug started in first line regimen by threshold windows.
Anchor drugs
EVG=Elvitegravir; DTG=dolutegravir; RAL=raltegravir;; ATV=atazanavir
LPV=lopinavir;; RPV=rilpivirine; EFV=efavirenz
NRTIs
TDF=tenofovir; 3TC=lamivudine; ZDV=zidovudine; FTC=emtricitabine; ABC=abacavir
NB only interpretations for the drugs included in first line regimen are shown in the graphs
At the >2%, T215 revertant (n=47, 3.4%), M41L (n=31, 2.3%), and K219QENR (n=28, 2.0%) were the most prevalent NRTI mutations. Mutations D67NGE and K70RE were detected in 1.2% of participants (n=17 and n=16, respectively). Notably, the K65R mutation was not observed in any sample, even at this low threshold. M184V and M184I were detected in 7 (0.5%) and 11 (0.8%) participants, respectively. The most common NNRTI DRMs were K103NS (n=47, 3.4%), G190ASE (n=39, 2.8%), and E138K (n=22, 1.6%). The prevalence of PI PDR was strongly influenced by the M46IL mutation, which was observed in 6.2% of participants (n=86). The D30N mutation was detected in 28 (2.0%) and the L90M in 6 (0.4%). Other commonly detected mutations in the PI region were the F53LY, I54VLMATS and V82ATFSCML (in approximately 0.6% of participants). The most common individual INSTI DRMs were T66AIK (n=6, 0.4%) G140ACRS (n=8, 0.6%) Y143CHR (n=7, 0.5%) Q148HKR (n=7, 0.5%) all with a prevalence <1%.
Overall, 85 participants met our definition of primary endpoint of treatment failure (TF). Using the 20% threshold, by two years from starting ART, the estimated proportion of participants who experienced TF were 13.0% (95%CI: 5.8–20.2) in the GSS 0–2.75 group vs. 5.0% (95% CI:3.8–6.3) in the GSS=3 group (log-rank p=0.006, Figure 2A ). Importantly, approximately 10% in the GSS 0–2.75 group (95%CI: 3.9–16.5%) experienced treatment failure already after 12 months of starting therapy.
Figure 2. Kaplan-Meier estimates of the time to treatment failure according to HIVDB GSS categories and threshold window.
Interestingly in the unweighted survival analysis, performed separately using the three threshold windows, the strongest evidence against the null hypothesis was for the Sanger threshold (>20%) as compared to the minority variants thresholds (Figures 2 A,B,C). For the >2% threshold, the Kaplan-Meier estimates of TF by 2.5 years were 7.1% (95% CI:3.2–10.9%) in the GSS 0–2.75 group vs. 5.0% (95%CI: 3.5–6.5) in the GSS=3 group (log-rank p=0.03, Figure 2C ).
The results obtained in the unadjusted analysis were confirmed after controlling for the set of identified potential confounders (Supplementary Figure S1, Table 2). In particular, we estimated a >2-fold difference in risk of TF by GSS group using the Sanger window threshold, although an effect as small as a 16% increase in risk as well as a 4-fold increase in risk were all values compatible with the data (Table 2, panel A). Of note, participants with a GSS in the 0–2.75 group were also at increased risk of TF when using the >2% window threshold for resistance, although the effect size was much smaller (adjusted for the confounders described in the DAG (Figure S1), HR=1.66, 95% CI: 1.01–2.71, p=0.04). Among the 27 ART discontinuations, the breakdown of the reason for the change were the following: 12 (44%) for failure (high HIV-RNA n=9, low CD4 count n=1 and detection of resistance n=2, one of which also had an elevated HIV-RNA), 9 (33%) for intolerance/toxicity, 4 (15%) for simplification (i.e. stop before switching to a simpler regimen) and the remaining 2 (0.7%) for other/unknown reasons. Results were similar when we used the alternative endpoint of treatment failure which only counted as events the discontinuations due to failure (n=77 total events, Table S2).
Table 2.
HR from fitting an unweighted Cox regression model with time fixed covariates at entry
| Panel A) | Unadjusted and adjusted hazard ratio of treatment failure endpoint& | |||
|---|---|---|---|---|
| Unadjusted HR (95% CI) | p-value | Adjusted* HR (95% CI) | p-value | |
| GSS (>20% window) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 2.29 (1.24, 4.21) | 0.008 | 2.18 (1.16, 4.09) | 0.015 |
| GSS (>5% window) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 1.91 (1.10, 3.34) | 0.022 | 1.74 (0.99, 3.07) | 0.056 |
| GSS (5–20% window £ ) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 1.72 (0.95, 3.10) | 0.074 | 1.56 (0.85, 2.84) | 0.150 |
| GSS (>2% window) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 1.69 (1.04, 2.75) | 0.033 | 1.66 (1.01, 2.74) | 0.046 |
| GSS (2–20% window £ ) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 1.57 (0.95, 2.60) | 0.079 | 1.54 (0.93, 2.56) | 0.096 |
| Panel B) | Subsets of participants who started a NNRTI-based regimen | |||
| GSS (>20% window) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 3.12 (1.61, 6.05) | <.001 | 2.93 (1.47, 5.84) | 0.002 |
| GSS (>5% window) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 2.58 (1.38, 4.81) | 0.003 | 2.25 (1.19, 4.25) | 0.012 |
| GSS (5–20% window £ ) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 2.27 (1.17, 4.42) | 0.016 | 1.99 (1.01, 3.90) | 0.046 |
| GSS (>2% window) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 2.35 (1.35, 4.12) | 0.003 | 2.32 (1.32, 4.09) | 0.003 |
| GSS (2–20% window £ ) | ||||
| 3 | 1 | 1 | ||
| 0–2.75 | 2.16 (1.21, 3.85) | 0.009 | 2.15 (1.20, 3.86) | 0.011 |
adjusted for geographical region, baseline HIV-RNA, intervention arm of START (immediate vs. differed) and year of starting ART
Pure confirmed VF or a single VL>200 copies/mL followed by a ART change
Restricting to those with GSS
≥2; this was done to exclude those who could be detected as participants initiating a suboptimal regimen by Sanger sequence alone
HIVDB v8.6 with >20% threshold
Interestingly in the main adjusted analysis, results were compatible with the null hypothesis of no difference when using the intermediate >5% window threshold. Of note, all these associations were much stronger and were all significant (p<0.05) when restricting the analysis to participants who started a regimen including an NNRTI as the anchor drug (Table 2, panel B, Table S3). The results of the sensitivity analysis after restricting to participants with 2–20% and 5–20% windows were consistent with those of the main analysis (Table 2A,B). Supplementary Figure S3 as well some supplemental text show the results of the analysis investigating the association between individual mutations and HIV-RNA decline.
DISCUSSION
One key finding of this analysis is the fact that pre-existing resistance detected in >20% of the virus population was a confirmed determinant of failure of first-line ART. Of note, participants in START were treatment-naïve as this was an entry criterion for the trial. Therefore, the most likely mechanism for study participants to have detectable HIV genotypic resistance at study entry would be due to infection with a resistant strain. Importantly, the results also confirm that PDR detected at >2% was also predictive of treatment failure in our study population. We estimated a 66% increase in risk of TF in those with suboptimal vs. optimal GSS but an increase risk as small as 1% was also compatible with the data, as indicated by the confidence limits. Importantly, the association was particularly strong when restricting to the participants who started first line therapy with NNRTI-based regimens (>2-fold increase in risk). The results were also confirmed in a number of sensitivity analyses.
Of interest, the proportion who started a regimen predicted by the Stanford GSS to have suboptimal activity was low at 6.7% (96/138) when using the routine detection threshold of >20%. This could be due to the lack of routine testing in a particular region of recruitment in the trial, the use of different systems to interpret the results or to the time between the date of the stored sample and that of ART initiation. Indeed, our data support the fact that people who started suboptimal therapy were enriched in South America as compared to Europe or the USA. The GSS 0–2.75 group also appeared to have their HIV infection diagnosed more recently than the GSS=3 group. In contrast, the median time from the date of stored sample and ART initiation was of 12 months due to the inclusion of participants in both study treatment arms, with no difference by GSS groups.
The analysis also confirms the results presented in our previous work, showing that a large proportion of the PDR mutations were detected at a level between 2% and 5% [24]. This appeared to be particularly marked for INSTI-associated mutations but to a lesser extent for NNRTI-associated mutations, namely rilpivirine. When the association between individual mutations and week 4 change in viral load was investigated, INSTI-associated mutations N155H and G140ACRS detected at the 2–20% threshold appeared to be those more strongly associated with a slower decline (association was not significant but prevalence of these mutations was very low impacting on the power of the analysis).
In a systematic review of 25 studies examining the impact of minority variants on initial ART, only 11 (44%) showed an association between the detection of pre-existing minority variants and risk of virological failure of NNRTI-based first line regimens [13,14, 22, 31–34]. However, the review included heterogeneous studies with respect of target population, definition of the exposure, exact regimen initiated and definition of the outcome and a quantitative meta-analysis has not been performed. Also, little was done to rank the studies according to quality of the study design and statistical methods employed (i.e. how confounding was handled, the presence of other systematic biases such as selection and observer bias). For these reasons, it is difficult to compare results and it appears to be misleading to present the crude percentage of studies in favour or against an association. Interestingly, the majority of studies showing no association were more recent analyses, conducted in the resource limited settings in people receiving RPV-based regimens [35–39]. Furthermore, the authors of this review, surprisingly, eventually conclude that minority variants have been shown to be clinically relevant, especially prior to the initiation of a first-line NNRTI-based regimen [22]. Our data support this conclusion.
One unmet need only partially addressed by this analysis is whether the data support a specific threshold for minority variants (i.e. 2% vs. 5% of virus populations) to be used in routine care.
Our findings seem to slightly favour the >2% threshold although much larger studies are needed stratifying by specific target populations, mutations and treatment started to be able to give recommendations for one threshold instead of another.
A number of limitations need to be discussed. One key limitation is the fact that the majority of participants initiated regimens that are no longer routinely started as first-line. Unfortunately, although one of the goals was to assess the role of minority variants to predict response to INSTI-based regimens, our analysis was underpowered to answer this question. Nevertheless, results are relevant for countries in which NNRTI-based regimens are still a prevalent option for first line therapy. WHO antiretroviral guidelines recommend the use of efavirenz as an alternative option in first-line ART regimens [40], which is still implemented in low- and medium-income counties (LMICs) if levels of pretreatment drug resistance to NNRTIs are <10%. Of course, all our results are valid under the set of strong, mainly untestable, assumptions, including no unmeasured confounding which we cannot rule out. Of note, by decreasing the threshold of the detection to 2% the number of unusual mutations increases which may lead to the detection of spurious associations because of misclassification of the exposure. Nevertheless, the NGS data were carefully reviewed by stipulating a minimum read depth of 200 across the HIV regions spanning all relevant mutations within each gene to minimise chance detection of unusual mutations at low levels. Finally, we did not have genotypic resistance tests done at time of failure. These additional data could have been useful to verify the extent of outgrowth of the minority variants detected at baseline and provide additional evidence towards a putative causative effect of minority variants as a key determinant of treatment failure. Also, it would have given an estimate of the potential impact of accumulated resistance on the response to second-line regimens.
In conclusion, this analysis confirms an association between detectable minority variants and risk of TF to first-line ART. Importantly, given prior conflicting results regarding the impact of NNRTI minority variants, the data confirms a strong association when the analysis was restricted to the subset of participants who initiated a NNRTI-based regimen. Further studies are needed to address other relevant unanswered questions, such as the role of minority variant INSTI-associated mutations and whether the ability of minority variants to predict treatment failure might vary by mutational load.
Supplementary Material
Acknowledgements
We wish to thank the study participants and clinical staff of the START trial (see Initiation of antiretroviral therapy in early asymptomatic individuals, New England Journal of Medicine 2015:373:794–807 for a complete list of START investigators [23]).
Financial disclosure
The study was supported in part by the National Institute of Allergy and Infectious Diseases (NIAID), National Institute of Health (NIH, grants UM1-AI068641, UM1-AI120197, and NHLBI grant RO1HL096453), National Institutes of Health Clinical Center, National Cancer Institute, National Heart, Lung, and Blood Institute, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institute of Mental Health, National Institute of Neurological Disorders and Stroke, National Institute of Arthritis and Musculoskeletal and Skin Diseases, Agence Nationale de Recherches sur le SIDA et les Hepatites Virales (France), National Health and Medical Research Council (Australia), National Research Foundation (Denmark), Bundes ministerium fur Bildung und Forschung (Germany), European AIDS Treatment Network, Medical Research Council (UK), National Institute for Health Research, National Health Service (UK), and University of Minnesota. Antiretroviral drugs were donated to the central drug repository by AbbVie, Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline/ViiV Healthcare, Janssen Scientific Affairs, and Merck
Footnotes
ACL declares no conflicts of interest.
REFERENCES
- 1.Stella-Ascariz N, Arribas JR, Paredes R, Li JZ. The Role of HIV-1 Drug-Resistant Minority Variants in Treatment Failure. J Infect Dis. 2017. Dec 1;216(suppl_9):S847–S850. doi: 10.1093/infdis/jix430. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Johnson JA, Li JF, Wei X, Lipscomb J, Irlbeck D, Craig C, Smith A, Bennett DE, Monsour M, Sandstrom P, Lanier ER, Heneine W. Minority HIV-1 drug resistance mutations are present in antiretroviral treatment-naïve populations and associate with reduced treatment efficacy. PLoS Med. 2008. Jul 29;5(7):e158. doi: 10.1371/journal.pmed.0050158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Li JZ, Kuritzkes DR. Clinical implications of HIV-1 minority variants. Clin Infect Dis. 2013. Jun;56(11):1667–74. doi: 10.1093/cid/cit125. Epub 2013 Feb 27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Paredes R, Marconi VC, Campbell TB, Kuritzkes DR. Systematic evaluation of allele-specific real-time PCR for the detection of minor HIV-1 variants with pol and env resistance mutations. J Virol Methods 2007; 146:136–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ellis GM, Vlaskin TA, Koth A, et al. Simultaneous and sensitive detection of human immunodeficiency virus type 1 (HIV) drug resistant genotypes by multiplex oligonucleotide ligation assay. J Virol Methods 2013; 192:39–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Jakobsen MR, Tolstrup M, Søgaard OS, et al. Transmission of HIV-1 drug-resistant variants: prevalence and effect on treatment outcome. Clin Infect Dis 2010; 50:566–73. [DOI] [PubMed] [Google Scholar]
- 7.Li JZ, Chapman B, Charlebois P, et al. ; ACTG A5262 Study Team. Comparison of Illumina and 454 deep sequencing in participants failing raltegravir-based antiretroviral therapy. PLoS One 2014; 9:e90485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gibson RM, Meyer AM, Winner D, et al. Sensitive deep-sequencing-based HIV-1 genotyping assay to simultaneously determine susceptibility to protease, reverse transcriptase, integrase, and maturation inhibitors, as well as HIV-1 coreceptor tropism. Antimicrob Agents Chemother 2014; 58:2167–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.2022. Update of the Drug Resistance Mutations in HIV-1 Topics in Antiviral Medicine. Volume 30 Issue 4 October/November 2022] [PMC free article] [PubMed] [Google Scholar]
- 10.Tang MW, Shafer RW. HIV-1 Antiretroviral resistance scientific principles and clinical applications. Drugs 2012; 72:1–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gibson RM, Weber J, Winner D, Miller MD, Quiñones-Mateu ME. Contribution of human immunodeficiency virus type 1 minority variants to reduced drug susceptibility in patients on an integrase strand transfer inhibitor-based therapy. PLoS One. 2014. Aug 11;9(8):e104512. doi: 10.1371/journal.pone.0104512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Fonager J, Larsson JT, Hussing C, Neess Engsig F, Nielsen C, Fischer TK. Identification of minority resistance mutations in the HIV-1 integrase coding region using next generation sequencing. J Clin Virol 2015; 73:95–100. [DOI] [PubMed] [Google Scholar]
- 13.Li JZ, Paredes R, Ribaudo HJ, et al. Low-frequency HIV-1 drug resistance mutations and risk of NNRTI-based antiretroviral treatment failure: a systematic review and pooled analysis. JAMA 2011; 305:1327–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cozzi-Lepri A, Noguera-Julian M, Di Giallonardo F, et al. ; CHAIN Minority HIV-1 Variants Working Group. Low-frequency drug-resistant HIV-1 and risk of virological failure to first-line NNRTI-based ART: a multicohort European case-control study using centralized ultrasensitive 454 pyrosequencing. J Antimicrob Chemother 2015; 70:930–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Halvas EK, Wiegand A, Boltz VF, et al. Low frequency nonnucleoside reverse-transcriptase inhibitor-resistant variants contribute to failure of efavirenz-containing regimens in treatment-experienced patients. J Infect Dis 2010; 201:672–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Casadellà M, Manzardo C, Noguera-Julian M, et al. ; ADVANZ and ADVANZ-3 Investigators. Clinical value of ultradeep HIV-1 genotyping and tropism testing in late presenters with advanced disease. AIDS 2015; 29:1493–504. [DOI] [PubMed] [Google Scholar]
- 17.Casadellà M, Noguera-Julian M, Sunpath H, et al. Treatment options after virological failure of first-line tenofovir-based regimens in South Africa: an analysis by deep sequencing. AIDS 2016; 30:1137–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Charpentier C, Lee GQ, Rodriguez C, et al. Highly frequent HIV-1 minority resistant variants at baseline of the ANRS 139 TRIO trial had a limited impact on virological response. J Antimicrob Chemother 2015; 70:2090–6. [DOI] [PubMed] [Google Scholar]
- 19.Kyeyune F, Gibson RM, Nankya I, et al. Low-frequency drug resistance in HIVinfected Ugandans on antiretroviral treatment is associated with regimen failure. Antimicrob Agents Chemother 2016; 60:3380–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Mohamed S, Penaranda G, Gonzalez D, et al. Comparison of ultra-deep versus Sanger sequencing detection of minority mutations on the HIV-1 drug resistance interpretations after virological failure. AIDS 2014; 28:1315–24. [DOI] [PubMed] [Google Scholar]
- 21.Nishizawa M, Matsuda M, Hattori J, et al. Longitudinal detection and persistence of minority drug-resistant populations and their effect on salvage therapy. PLoS One 2015; 10:1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Mbunkah HA, Bertagnolio S, Hamers RL, et al. Low-Abundance Drug-Resistant HIV-1 Variants in Antiretroviral Drug-Naive Individuals: A Systematic Review of Detection Methods, Prevalence, and Clinical Impact. The Journal of infectious diseases 2020; 221(10): 1584–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Group ISS, Lundgren JD, Babiker AG, et al. Initiation of Antiretroviral Therapy in Early Asymptomatic HIV Infection. The New England journal of medicine. 2015; 373:795–807. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Baxter JD, Dunn D, Tostevin A, Marvig RL, Bennedbaek M, Cozzi-Lepri A, Sharma S, Kozal MJ, Gompels M, Pinto AN, Lundgren J; INSIGHT START Study Group. Transmitted HIV-1 drug resistance in a large international cohort using next-generation sequencing: results from the Strategic Timing of Antiretroviral Treatment (START) study. HIV Med. 2020. Dec 25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Marc Bennedbæk Anna Zhukova, Man-Hung Eric Tang Jaclyn Bennet, Munderi Paula, Ruxrungtham Kiat, Gisslen Magnus, Worobey Michael, Lundgren Jens D, Marvig Rasmus L, for the INSIGHT START study group, Phylogenetic Analysis of HIV-1 Shows Frequent Cross-Country Transmission and Local Population Expansions, Virus Evolution, 2021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Verbist B, Thys K, Reumers J, et al. VirVarSeq: a low-frequency virus variant detection pipeline for Illumina sequencing using adaptive base-calling accuracy filtering. Bioinformatics (Oxford, England). 2015; 31:94. [DOI] [PubMed] [Google Scholar]
- 27.Baxter J, Dunn D, White E, et al. Global HIV-1 transmitted drug resistance in the INSIGHT S trategic T iming of A nti R etroviral T reatment (START) trial. HIV medicine. 2015; 16:77–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bennett DE, Camacho RJ, Otelea D, et al. Drug resistance mutations for surveillance of transmitted HIV-1 drug-resistance: 2009. update. PloS one. 2009; 4:e4724. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Database SUHDR. 2016. Surveillance Drug Resistance Mutation (SDRM) Worksheet:INIs. [Google Scholar]
- 30.Liu TF and Shafer RW. Web resources for HIV type 1 genotypic-resistance test interpretation. Clinical infectious diseases. 2006; 42:1608–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Inzaule SC, Hamers RL, Noguera-Julian M, et al. ; PanAfrican Studies to Evaluate Resistance. Clinically relevant thresholds for ultrasensitive HIV drug resistance testing: a multicountry nested case-control study. Lancet HIV 2018; 5:e638–46. [DOI] [PubMed] [Google Scholar]
- 32.Geretti AM, Fox ZV, Booth CL, et al. Low-frequency K103N strengthens the impact of transmitted drug resistance on virologic responses to first-line efavirenz or nevirapinebased highly active antiretroviral therapy. J Acquir Immune Defic Syndr 2009; 52:569–73. [DOI] [PubMed] [Google Scholar]
- 33.Metzner KJ, Giulieri SG, Knoepfel SA, Rauch P, Burgisser P, Yerly S, Günthard HF, Cavassini M. Minority quasispecies of drug-resistant HIV-1 that lead to early therapy failure in treatment-naive and -adherent patients. Clin Infect Dis. 2009. Jan 15;48(2):239–47. [DOI] [PubMed] [Google Scholar]
- 34.Paredes R, Lalama CM, Ribaudo HJ, Schackman BR, Shikuma C, Giguel F, Meyer WA 3rd, Johnson VA, Fiscus SA, D’Aquila RT, Gulick RM, Kuritzkes DR; AIDS Clinical Trials Group (ACTG) A5095 Study Team. Pre-existing minority drug-resistant HIV-1 variants, adherence, and risk of antiretroviral treatment failure. J Infect Dis. 2010. Mar;201(5):662–71. doi: 10.1086/650543. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Metzner KJ, Scherrer AU, von Wyl V, et al. ; Swiss HIV Cohort Study. Limited clinical benefit of minority K103N and Y181C-variant detection in addition to routine genotypic resistance testing in antiretroviral therapy-naive patients. AIDS 2014; 28:2231–9. [DOI] [PubMed] [Google Scholar]
- 36.Raymond S, Nicot F, Pallier C, et al. ; French National Agency for Research on AIDS and Viral Hepatitis (ANRS) AC11 Resistance Study Group. Impact of human immunodeficiency virus type 1 minority variants on the virus response to a rilpivirine-based first-line regimen. Clin Infect Dis 2018; 66:1588–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Su B, Zheng X, Liu Y, et al. Detection of pretreatment minority HIV-1 reverse transcriptase inhibitor-resistant variants by ultra-deep sequencing has a limited impact on virological outcomes. J Antimicrob Chemother 2019; 74:1408–16. [DOI] [PubMed] [Google Scholar]
- 38.Van Eygen V, Thys K, Van Hove C, et al. Deep sequencing analysis of HIV-1 reverse transcriptase at baseline and time of failure in patients receiving rilpivirine in the phase III studies ECHO and THRIVE. J Med Virol 2016; 88:798–806. [DOI] [PubMed] [Google Scholar]
- 39.Zoufaly A, Jochum J, Hammerl R, et al. Virological failure after 1 year of first-line ART is not associated with HIV minority drug resistance in rural Cameroon. J Antimicrob Chemother 2015; 70:922–5. [DOI] [PubMed] [Google Scholar]
- 40. https://www.who.int/publications/i/item/9789240031593 .
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.