We should not stop considering HIV drug resistance testing at failure of first-line antiretroviral therapy

Abstract

HIV drug resistance is a major global hurdle to successful and sustained antiretroviral therapy. Global guidelines recommend testing for antiretroviral drug resistance and results are used to inform treatment regimen design for patients at different stages of therapy. Several clinical trials investigated optimal regimens after failure of first-line antiretroviral therapy, yielding data that advanced knowledge and informed care. However, further interpretation of data from these studies questioned the benefit of antiretroviral drug resistance testing for cases in which first-line treatment is not effective and, furthermore, that relying on the results of antiretroviral drug resistance testing could be misleading and unnecessary. In this Viewpoint, which is largely focused on high-income settings, we review these data, reflect on the potential problems with their interpretation, and call for caution in their extrapolation. Without negating the importance of the data, and recognising the varied circumstances related to HIV drug resistance testing in different global settings, we advise caution before changing current practice and recommendations. We believe that we should not universally stop considering HIV drug resistance testing at failure of first-line antiretroviral therapy.

Introduction

Advances in antiretroviral therapy have substantially improved treatment outcomes for people with HIV. Increased global access and improved formulations with better toxicity profiles and higher barriers to drug resistance mean that HIV drugs are now easier to obtain, more tolerable, and less likely to result in drug resistance than in the past.^1,2 However, drug resistance, which has been a cause and consequence of antiretroviral treatment failure since the introduction of zidovudine in the 1980s, remains a major hurdle to sustained lifelong therapy.³

Testing for resistance to antiretroviral drugs and use of the results in regimen design is recommended in guidelines for individual care if feasible.⁴ In settings where resistance testing is not feasible (eg, in some resource-limited settings) HIV treatment guidelines recommend more targeted approaches on the basis of age or treatment experience.^5,6 Cost and infrastructure constraints are the primary factors limiting access to HIV drug resistance testing (DRT), and where it is unavailable testing is sought after.^7–9 Specific mutations detected in the HIV genome predict antiretroviral susceptibility and inform care providers in designing subsequent successful regimens.¹⁰ Although this approach is not yet practicable in the most resource-limited settings for individual care, WHO recommends surveillance and monitoring of antiretroviral drug resistance at a population level in these settings.¹¹ Action to expand, optimise, and economise HIV DRT is also encouraging.¹²

We selected several trials (according to our expertise and judgement) that investigated regimen design after failure of first-line antiretroviral therapy, yielding data that advance knowledge and help to inform care. However, further interpretation of these data, as presented and discussed in the reports of these trials, suggests little benefit from DRT when first-line antiretroviral therapy has failed, and that relying on such data may be misleading and thus unnecessary. In this Viewpoint, we reflect on these interpretations of trial data, and call for more caution in their extrapolation. We note two important points: first, currently this discussion is mostly relevant for high-resource settings, where recommendations for DRT on first-line antiretroviral failure are already incorporated into guidelines and practised in clinical care; and second, the goal of this Viewpoint is not to try to re-visit whether DRT on first-line antiretroviral failure should be considered, but rather to alert that available data do not convincingly make a case for change.

We identified six studies (table 1) that explored second-line treatment regimens with or without DRT on failure of first-line antiretroviral therapy: EARNEST,^13,14 SECOND-LINE,^15,16 SELECT,¹⁷ DAWNING,¹⁸ NADIA,^19,20 and REVAMP.²¹ Collectively, data support the use of a boosted protease inhibitor (lopinavir or darunavir) or an integrase strand transfer inhibitor (INSTI; raltegravir or dolutegravir) together with two nucleoside reverse transcriptase inhibitors (NRTIs; lamivudine, or emtricitabine with tenofovir or zidovudine), and NRTI-sparing regimens that use a combination of INSTI and protease inhibitors as second-line therapy after non-nucleoside reverse transcriptase inhibitor (NNRTI)-based first-line antiretroviral therapy.

Table 1:

Summary of studies considered in this Viewpoint

	Setting	Follow-up	Failing first-line ART	Viral load cutofffor DRT (copies per mL)	DRT used forswitch	Resista nce at first-line failure	Study arms	Cutoff for VF (copies per mL)	Virological outcome	DR outcomes	Comments and conclusions
EARNEST13 (n=1277)	Kenya, Malawi, Uganda, Zambia, Zimbabwe	96 weeks	NNRTI-based (69% zidovudine, 62% stavudine, 12%tenofovir disoproxil fumarate)	>1000	No	57% intermediate to high predicted resistanceto tenofovir disoproxil fumarate	Arm 1 LPV/r + 2–3 NRTIs; arm 2 LPV/r + raltegravir; arm 3 LPV/r (inferior)	400	Arm 186% VS; arm 2 86% VS; arm 3 6l%VS	Arm 1 LPV/r intermediate to high predicted resistance 2%, NRTIs intermediate to high predicted resistance 4% (excluding lamivudine or emtricitabine); arm 2 LPV/r intermediate to high predicted resistance 1%, raltegravir 3%; arm 3 LPV/r intermediate to high predicted resistance 18%, DRV/r 4% intermediate predicted resistance, 0% high predicted resistance	With PI, NRTIs retain substantial activity; no advantageto replace NRTIs with raltegravir; residual NRTI activity despite resistance
EARNEST14 (n=391)	Kenya, Malawi, Uganda, Zambia, Zimbabwe	144 weeks	NNRTI-based (69% zidovudine, 62% stavudine, 12%tenofovir disoproxil fumarate)	>1000	No	57% intermediate to high predicted resistanceto tenofovir disoproxil fumarate	Arm 1 LPV/r + 2–3 NRTIs; arm 2 LPV/r + raltegravlr; arm 3 LPV/r	400	Arm 18 9% VS with no active NRTIs; arm 2 85% VS with 1 active NRTI; arm 3 77% VS with 2–3 active NRTIs	..	Greater predicted NRTI activity associated with worse suppression, probably due to adherence; DRT might overestimate cross-resistance; results do not support introduction of routine DRT to select second-line NRTIs In resource-limited settings
SECOND- LINE15(n=541)	Global	48 weeks	NNRTI-based	>500	Forsome	98% >1NRTI or NNRTI mutation	Arm 1 LPV/r + 2–3 NRTIs; arm 2 LPV/r + raltegravir (non-inferior)	200	Armi 81% VS; arm 2 83% VS	Arm 16/43 (14%) VF with NRTI DR; arm 2 7/47 (15%) VF with raltegravir DR	NRTI-free strategy feasible; high suppression despite DR; no outcome difference if guided by genotype
SECOND- LINE16 (n=215)	Global	96 weeks	NNRTI-based first-line failure	>500	For some	Median gGSS 3 in the NRTI group and 3 In the raltegravir group; median sGSS 1 in the NRTI group	Armi LPV/r + 2–3 NRTIs; arm 2 LPV/r + raltegravlr	200	Armi 76% VS; arm 2 80% VS (non-inferior)	Emergent major mutations at VF 1% p rotease, 13% NRTIs, 20% integrase	VF associated with adherence; when limited resources invest in adherence optimisation not DRT; more mutations lead to better suppression; recycled NRTIs confer residual activity despite resistance predietion; DRT not justlfied for early or late stages
SELECT (A5273)17 (n=515)	Global (resource-limited settings)	48 weeks	NNRTI-based first-line failure	>400	No	Excluded challenging DR; 95% NRTI mutations; 48% NRTI GSS<1	Arm 1 LPV/r + raltegravir (non-inferior); arm 2 LPV/r + 2–3 NRTIs	400	Armi 90% VS; arm 2 88% VS	26% in raltegravir group with new DR; 29% in NRTI group with new DR	Good VS despite high NRTI DR; better VS with higher NRTI DR, probably adherence related; focus on optimisation of drug availability and adherence rather than DRT after first-line failure
DAWNING18 (n=968)	Global	48 weeks	NNRTI-based first-line fallure	Not defined	Yes	88% NRTI and NNRTI DR; all ≥1 active NRTI by DRT	Armi dolutegravlr + 2 NRTIs (≥1 fully active) (superior); arm 2 LPV/r + 2 NRTIs (≥1 fully active)	400	Armi 84%VS; arm 2 70% VS	Of 11 VF In dolutegravlr group, 2 with new dolutegravlr DR; of 30 VF In LPV/r group, 3 with new NRTI DR	BetterVS with hlgher NRTI DR
NADIA19 (n=464)	Kenya, Uganda, Zimbabwe	48 weeks	NNRTI +tenofovir disoproxil fumarate or lamivudine first-line failure	1000	No	59% intermediate to high predicted resistance to tenofovir disoproxil fumarate; 58% on tenofovir disoproxil fumarate with no active NRTIs	dolutegravir vs DRV/r; tenofovir disoproxil fumarate vs zidovudine; all + lamivudine	400	90% VS on dolutegravir; 92% VS on DRV/r; 92% VS on tenofovir disoproxil fumarate; 90% VS on zidovudine; 90% VS with no NRTI activity	4/14 (6%) with VF on dolutegravir with dolutegravir DR, 3 in zidovudine; 0/13 with VF on DRV/r with darunavir DR	Dolutegravir non-inferior to DRV/r; dolutegravir + 2 NRTIs effective, including with extensive NRTI DR; tenofovir disoproxil fumarate non-inferior to zidovudine; results generalisable to most patients globally with same HIV care; DRT algorithms may need revision
NADIA20 (n=464)	Kenya, Uganda, Zimbabwe	96 weeks	NNRTI +tenofovlr disoproxil fumarate or lamivudine first-line fallure	1000	No	59% intermediate to high predicted resistance tenofovir disoproxil fumarate DR; 58% on tenofovir disoproxil fumarate with no active NRTIs	Dolutegravlr vs DRV/r; tenofovir disoproxil fumarate vs zidovudine; all + lamivudine	400	90% VS on dolutegravlr; 87% VS on DRV/r; 92% VS on tenofovir disoproxil fumarate; 85% VS on zidovudine	9/20 (4%) with VF on dolutegravir; with dolutegravir DR (5 high predicted reslstance, 4 Intermediate predicted reslstance; 3 In tenofovir disoproxil fumarate, 6 In zidovudine); 0/26 with VF on DRV/rwIth darunavir DR	Non-inferiority of dolutegravir to DRV/r, but greater risk for resistance; superiority of tenofovir disoproxil fumarate over zidovudine; tenofovir disoproxil fumarate should be continued In second-line, rather than switch to zidovudine; VS possible even with NRTI DR;those with past VF and N RTI DR may be at higher risk of VF; findings also applicable to Individualised ART
REVAMP21	Uganda, South Africa	36 weeks	N N RTI - based first-1ine failure; 72%tenofovir disoproxil fumarate + emtricitabine + efavirenz	Notdefined	Inthe DRT arm	68% with DR	Arm 1 SOC (adherence, repeatVL); arm 2 DRT at first-line failure	200	Armi 61%VS; arm 2 75% VS; 75% VS in Uganda, 48% VS in South Africa	Among VF, DRT arm more likely to have no DR (76% vs 59%)	DRT does not improve resuppression; lessVF with DR in the DRT arm; results do not support DRT use in general

ART=antiretroviral therapy. DR=drug resistance. DRT=drug resistance testing. DRV/r=ritonavir-boosted darunavir. GSS=genotypic susceptibility score. gGSS=global genotypic susceptibility score. NNRTI=non-nucleoside reverse transcriptase inhibitor. NRTI=nucleoside or nucleotide reverse transcriptase inhibitor. LPV/r=ritonavir-boosted lopinavir. PI=protease inhibitor. sGSS,=specific genotypic susceptibility score. SOC=standard of care. VF=virological failure. VS=virological suppression

The reviewed studies also demonstrate that people with high levels of drug resistance at failure of first-line therapy have better viral suppression after switching to second-line antiretroviral therapy than do those without resistance; and that many people with HIV achieve virological suppression despite significant NRTI resistance, which predicts reduced or no action of certain drugs in second-line antiretroviral therapy (table 1). The aggregated deduction presented and discussed in these studies is that DRT is unnecessary on failure of first-line treatment; but we feel that interpretation of these data and extrapolation to policy and practice require careful consideration.

Considerations in data interpretation

Before adopting recommendations to avoid HIV DRT at failure of first-line antiretroviral therapy, there are a few matters to pause and consider. First, the studies do not provide individual data points to allow evaluation of the comprehensiveness and strength of this recommendation. Raw sequence data enabling examination of specific mutation patterns, and data linking these patterns to patients’ characteristics (eg, complete antiretroviral history or past viral loads), would enable differentiation between populations of people receiving therapy; for example, the simplicity or complexity of resistance mutation patterns might influence whether someone will benefit from DRT or not.^22,23 Of the studies reviewed, the SECOND-LINE study offers publicly available sequence data and therefore demonstrates the resistance mutation spectrum (table 2).^10,15 Although mutation patterns can be derived from sequences to demonstrate this wide spectrum, the unavailability of linked individual characteristics—such as complete treatment and viral load data, which were not linked to resistance data in the SECOND-LINE study—prevents further evaluation of potential benefits of DRT. Avoiding DRT for even some of these individuals might, in fact, worsen the problem and prevent design of more adequate regimens. However, raw sequence and linked data to substantiate or negate this statement are unavailable in the reviewed studies.

Table 2:

Spectrum of drug resistance at failure of first-line antiretroviral therapy in six patients from the SECOND-LINE study²⁴

	Previous antiretroviral therapy	Previous viral load detection	NRTI mutations at first-line viral failure	NNRTI mutations at first-line viral failure
Patient 1	NA	NA	None	None
Patient 2	NA	NA	Metl84Val	Lys103Asn
Patient 3	NA	NA	Lys65Arg	Val179Glu, Tyr181Cys, and Gly190Ala
Patient 4	NA	NA	Lys65Arg, Met184Val	Val106Ile, Val179Glu, and Tyr188Leu
Patient 5	NA	NA	Lys65Arg, Lys70Thr, Val75Met, Phe77Leu, Tyr115Phe, Phe116Tyr, Gln151Met, and Met184Val	Val179Glu, Tyr181Cys, and His221Tyr
Patient 6	NA	NA	Asp67Asn, Lys70Arg, Met184Val, Thr215Ile, and Lys219Gln	Lys101His, Lys103Asn, Val108Ile, Tyr181Cys, Gly190Ala, and His221Tyr

All study sequences were downloaded from the Stanford HIV Drug Resistance Database and the resistance mutation profiles of the six patients in the table were selected to represent the wide spectrum from no mutations to a large number of mutations, demonstrating the importance of examining mutations profiles and linking them to individual characteristics; however, no data on key characteristics of subsequent antiretroviral therapy, follow-up viral loads, or subsequent drug resistance testing were available. NA=not available. NNRTI=non-nucleoside reverse transcriptase inhibitor. NRTI=nucleoside reverse transcriptase inhibitor.

Second, beyond the relatively short follow-up periods of most reviewed studies, they also do not provide complete individual longitudinal data. Accumulation of drug resistance can affect regimen design for lifelong antiretroviral therapy. Investigating drug resistance evolution on an individual basis is crucial, yet usually unavailable and can potentially inform differential benefit of DRT. Some data are available. In NADIA, some information is provided for nine individuals in the group receiving dolutegravir plus two NRTIs, who had failure of second-line therapy with meaningful dolutegravir resistance. In the REVAMP trial, participants in the standard-of-care group were more likely to have resistance to their regimen upon failure of second-line treatment. These data are concerning, and the relationship to individuals’ drug resistance patterns at failure of first-line and second-line treatment, past treatment history, and viral load patterns would be important and informative when considering use of DRT at first-line treatment failure.

Third, resistance cannot be predicted for all antiretrovirals with similar accuracy; and even if full, partial, or no activity can be accurately predicted, the impact of resistance mutations on treatment outcomes depends on what alternative drugs and combinations are available. This uncertainty is particularly relevant for NRTIs, for which resistance interpretation algorithms and expert opinion can differ. This difference has relevance both for NNRTI-based regimens (as in the reviewed studies) and for newer regimens (eg, INSTI-based or dual-medication combinations), in which NRTIs are important components.²⁵ The spectrum of options is wide, and consideration of partial or no activity of some antiretrovirals in combination with full activity of others, as some of these studies suggest, can be challenging to interpret. DRT might therefore be beneficial to sort through these options, even if only in some instances, regimens, individuals, and settings; and even if currently some settings are not able to act on DRT results.

Fourth, many of the studies assume a public health approach, rather than focusing on the individualised care of people with HIV. This approach is reasonable and necessary; however, there are two potential risks in making conclusions and recommendations about DRT from a public health perspective. First, some settings do allow individual-level monitoring (including frequent viral load tests and DRT) and caution should be practised in extrapolating data derived from a public health approach onto these settings. Second, a recommendation that such testing is unnecessary carries a risk of disrupting innovations to make this important tool more widely available where it is not currently feasible.

Finally, several methodological limitations should be recognised as we contemplate how to interpret and whether to extrapolate data on drug resistance from the included studies. First, specific inclusion criteria lead to distinct populations in the studies defined by, for example, limited previous viral load suppression when on first-line NNRTI regimens, and exposure to previous antiretrovirals (eg, 6% in the NADIA trial had zidovudine therapy before therapy with tenofovir disoproxil fumarate).²⁶ Second, viral load cutoffs used to define treatment failure (eg, viral load <400 copies per mL in EARNEST and NADIA, and viral load <1000 copies per mL in REVAMP), and the concern for drug resistance below these thresholds could inform resistance interpretation and treatment options (eg, of the nine individuals with dolutegravir resistance in NADIA, two had viral loads between 400 and 1000 copies per mL). Third, whether DRT was done in real time to inform second-line regimen design (table 1) and, if so, which medication changes were made, could affect the usefulness and need for DRT. Fourth, specific methods used to evaluate resistance and predicted antiretroviral therapy activity influence generalisability (eg, in NADIA, specific NRTI resistance mutations were considered rather than predicted drug activity due to potential collinearity and complexity). Fifth, the potential for transmitted or pre-treatment NNRTI resistance to be associated with response to dolutegravir-based antiretroviral therapy can enhance the relevance of DRT.²⁷

Considerations in data extrapolation

Alternative observations can be made from the available data regarding DRT on failure of first-line antiretroviral therapy. First, understanding of DRT results relies on complex algorithms.¹⁰ These useful tools have been successfully applied in HIV care globally.²⁵ The knowledge base that enables such algorithms relies on expert interpretation of in vitro, genotypic-to-phenotypic, and clinical outcomes data.¹⁰ The limitations of such algorithms and the need to evaluate them continuously and rigorously have been recognised throughout the use of antiretroviral therapy.²⁵ Data discussed here are likely to provide further evidence of the imperfection of these algorithms, as also recognised by the NADIA authors.^19,20 For example, perhaps current prediction of intermediate resistance to tenofovir disoproxil fumarate in combination with other antiretrovirals on the basis of mutation patterns should be revised (eg, the common mutation Met184Val associated with resistance to lamivudine or emtricitabine might confer greater hyper-susceptibility, or the Lys65Arg mutation might confer less resistance to tenofovir disoproxil fumarate in particular combinations than currently recognised). The logical deduction here is that algorithms should be improved, not that DRT is unnecessary. However, without investigating individual-level data, such explorations are not feasible.

In studies with monotherapy arms (eg, EARNEST; table 1) some activity of NRTIs is likely to exist beyond the expected activity of the anchor drugs with a high barrier to resistance, such as boosted protease inhibitors and INSTIs. Beyond specific regimens used, monotherapy of such medications can sometimes be effective, although inadvisable.^4,28 Assessing its specific benefit in different mutations and therapeutic circumstances would be valuable and informative, to better understand the potential usefulness of DRT. Moreover, not all monotherapies are the same; dolutegravir monotherapy might be more challenging than boosted protease inhibitor monotherapy in terms of emergent resistance.^28,29 The logical deduction here is that perhaps the benefit observed in some of the studies resulted from monotherapy or partial monotherapy (ie, with other partly active medications), which DRT in a clinical setting can identify. Again, the scarcity of individual-patient data prevents such analyses.

A further topic worth investigating is HIV-1 subtype heterogeneity. Although subtype heterogeneity is an active research topic, the relevance to clinical care and the development of drug resistance is still unclear. Perhaps other ecological factors should be examined as variables, and more research is needed into whether and how these considerations can be incorporated into algorithms.

Different recommendations on timing of DRT (before antiretroviral therapy, at first-line failure, and at later failures) might be confusing both in settings where DRT is available and in settings that strive to implement it. Whether these recommendations make biological or scientific sense should be explored before formulating policy.

Finally, laboratory tests rely on clinical judgement and expertise, and HIV DRT is no exception.³⁰ Multiple factors go into clinical interpretation of a drug resistance report, such as medication adherence, drug-to-drug interactions, comorbidities, past treatment failures, accumulated drug resistance, and patient preference. Some of these considerations have already been discussed, but specifics of mutation susceptibility and predicted susceptibility are also important. For example, HIV experts are likely to use resistance reports to design subsequent regimens, including consideration of medications with low, intermediate, or even high values for predicted resistance. Clinical judgement will be led by available antiretroviral therapy and its formulations (eg, single tablet regimens vs single drug pills). However, this is a separate consideration to whether DRT should be used in cases of first-line failure of antiretrovirals. The logical deduction here would be that HIV DRT is an added tool in the range of considerations for clinical care and regimen design, even if its accuracy might be less than perfect.

Recommendations

Our main goal in this Viewpoint is to caution the over-extrapolation of published data on the usefulness of HIV DRT at failure of first-line antiretroviral therapy. We note several concepts that come out of this discussion and that we think should be considered to guide further research as this topic continues to be debated. These specific recommendations can provide guidance when considering DRT in cases for which first-line antiretrovirals are ineffective, its incorporation in treatment guidelines, and design of further studies and analyses to inform related decisions. Individual-level data such as drug resistance profiles, antiretroviral therapy, viral load histories, and longitudinal outcomes should be considered to allow objective research interpretations. Additional non-individual-level variables that impact such interpretations and reporting should also be taken into account, including antiretroviral failure thresholds, drug resistance at low viral loads, the use of DRT to guide therapy, and how this guidance was used for regimen design. The remaining activity of each antiretroviral at first-line failure and the ability of DRT to guide their use in subsequent regimens, should be considered at the patient level. Training on interpretation of DRT results at failure of first-line antiretroviral therapy is needed to improve clinical judgement and expertise. Drug resistance interpretation algorithms should continue to be expanded to incorporate newly available and existing data. The potential relevance of subtype to interpretation of DRT should be researched further.

Conclusions

The studies considered here provide valuable data to inform HIV care. However, caution needs to be exercised before using the data to change the currently accepted framework and to suggest that DRT is not relevant for people with HIV at failure of first-line antiretroviral therapy. Rather, limitations should be recognised, and available data should be used to address reservations about the use of DRT. Caution should also be used in making wider extrapolations and interpretations of the data and the potential impact of these on resulting recommendations. The limitations of data and their effects on recommendations might be even more relevant and different with global changes in antiretroviral therapy (eg, INSTIs replacing NNRTIs and their incorporation into all prevention and treatment modalities, and two-drug and long-acting regimens for treatment and prevention), and increasingly better, simpler, and cheaper treatment monitoring availability. This discussion is currently more relevant for high-income settings, where DRT in patients with first-line treatment failure is already incorporated into guidelines and used in clinical practice. However, use of DRT for individual clinical care in low-income and middle-income countries should not be overlooked as we work towards infrastructure and cost improvements to enable the consideration of this important tool in all clinical settings. We encourage careful examination and critical interpretation of data by providers, researchers, and policy makers, to continue to ensure the best outcomes of HIV care and further reductions in global HIV morbidity, mortality, and transmission. For now, we believe that we should not universally stop considering HIV DRT at failure of first-line antiretroviral therapy.