HIV-1 Drug Resistance Trends in the Era of Modern Antiretrovirals: 2018–2024

Abstract

Background

Antiretroviral drug resistance limits treatment options for people with human immunodeficiency virus (HIV) 1 and may reduce the effectiveness of preexposure prophylaxis. Novel treatment options with enhanced efficacy and more convenient formulations have become available from 2016 to 2021. Large-scale studies of trends in the prevalences of plasma RNA drug resistance mutations (DRMs) since 2018 are lacking, and there have been no systematic studies of trends in proviral DNA DRMs.

Methods

We retrospectively analyzed deidentified HIV-1 plasma RNA and proviral DNA sequences from specimens submitted to a reference laboratory between January 2018 and May 2024. We analyzed the annual prevalence of DRMs with a Stanford HIV Drug Resistance Database score of ≥30 for nucleoside and nonnucleoside reverse-transcriptase inhibitors (NRTIs and NNRTIs), protease inhibitors, and integrase strand transfer inhibitors (INSTIs).

Results

The prevalence of resistance declined for both RNA and DNA sequences. Single-class and dual-class NRTI + NNRTI resistance declined but was higher for DNA (NRTI + NNRTI, declined from 6.1% to 3.5% for RNA and from 12.1% to 7.8% for DNA). Rilpivirine DRMs remained low, with prevalences of 6.3% (RNA) and 10.2% (DNA) in 2024. The doravirine DRM prevalences in 2024 were 2% (RNA) and 2.9% (DNA). INSTI and dual-class NRTI + INSTI resistance also declined, but the prevalence of integrase DRM R263K increased.

Conclusions

Prevalence of NRTI and NNRTI resistance has declined, consistent with increased use of regimens with higher resistance barriers, improved tolerability, and more convenient dosing. Proviral DNA resistance trends were correlated with those for RNA. Continued advances in antiretroviral therapy efficacy, durability, and tolerability may lead to increased rates of virologic suppression and further reduce the incidence of archived resistance mutations in proviral DNA.

Using a large, real-world reference laboratory database, we demonstrated that human immunodeficiency virus 1 antiretroviral drug resistance has declined in plasma RNA and proviral DNA, consistent with increased use of regimens with higher resistance barriers, improved tolerability, and more convenient dosing.

In 2023 there were an estimated 39.9 million people with human immunodeficiency virus (HIV; PWH) globally, with 1.3 million new HIV infections yearly [1]. In the United States, the incidence of new HIV infections in 2022 was 31 800, and the overall number of PWH ≥13 years of age was estimated at 1 238 000 [2]. Antiretroviral (ARV) therapy (ART) has dramatically increased survival in PWH and reduced morbidity and mortality rates [3, 4]. It has also proved highly effective in preventing the transmission of HIV infection in individuals with virologic suppression [5, 6]. While approximately 77% of PWH have access to ART [1], in 2022 only 65% of PWH in the United States achieved virologic suppression [7]. Acquired and transmitted drug resistance remain obstacles to successful ART and to ending the HIV epidemic as a public health threat by 2030 [8–11].

For individuals receiving an ART regimen with a high barrier to resistance, nonadherence, pharmacologic factors, or transmitted resistant variants may lead to the emergence of drug resistance mutations (DRMs) [12–15]. Prior treatment failure, high pre-ART viral RNA levels, and low pre-ART CD4 cell counts may also increase the risk of treatment failure [16]. The acquisition of DRMs may lead to the loss of virologic suppression, limit future treatment options, and reduce the effectiveness of preexposure prophylaxis (PrEP) in preventing the transmission of HIV. Plasma RNA genotypic resistance testing (GRT) of HIV-1 to detect DRMs associated with resistance to nucleoside reverse-transcriptase (RT) inhibitors (NRTIs), nonnucleoside RT inhibitors (NNRTIs), and protease (PR) inhibitors (PIs) has been the standard of care for >25 years and is recommended by clinical practice guidelines [17, 18]. In addition, GRT to detect integrase strand transfer inhibitor (INSTI) DRMs is recommended if transmitted integrase resistance is suspected, in cases of treatment failure on an INSTI regimen. and for individuals who contracted HIV while receiving cabotegravir for PrEP [18]. INSTI GRT has been commercially available in the United States since 2008 [17, 18].

Prior studies using large commercial reference laboratory databases have documented a decline in single-class and multiclass drug resistance before 2018 [19–21]. However, a number of new ARVs have entered into clinical use since 2018, including the INSTI bictegravir, the NNRTI doravirine, and a long-acting rilpivirine and cabotegravir injectable regimen. In addition, proviral DNA resistance testing is increasingly being used in virologically suppressed patients before regimen switches. Large-scale studies of trends for plasma RNA DRM prevalence since 2018 are lacking, and to our knowledge, there have been no systematic studies of trends in proviral DNA DRMs. This study used a large and representative reference laboratory database encompassing aggregated deidentified HIV-1 RNA and DNA sequences to assess US trends in resistance from 2018 to 2024.

METHODS

A database of deidentified HIV-1 PR, RT, and integrase sequences was derived from specimens submitted for routine plasma RNA (>90 000 specimens) or proviral DNA GRT (>25 000 specimens) between January 2018 and May 2024. Data were limited to US PWH between the ages of 1 and 90 years, and multiple GRT for an individual placed within 1 day were tabulated only once. To tabulate multicategory drug resistance (NRTI + NNRTI, NRTI + INSTI, and NRTI + NNRTI + INSTI), plasma RNA sequences were further limited to those with RT, PR, and integrase GRT ordered concurrently. Demographic data were limited to sex (male, female, or unknown), age, and US Department of Health and Human Services (HHS) region [22]. This study was conducted with aggregated deidentified data from reference laboratory testing and did not include factors necessitating patient consent.

Plasma RNA GRT used Sanger sequencing, and bidirectional reads were assembled to an HIV-1 subtype B consensus sequence. Proviral DNA GRT used next-generation sequencing (NGS) with a minority variant (MV) cutoff of 10% on the Illumina MiSeq platform. NGS reads were filtered for stop codons, frameshifts, large deletions, enzymatic active site mutations, and excessive APOBEC-context mutations using the Hypermut 2.0 algorithm [23]. These likely defective viral reads were removed before mapping to an HXB-2 HIV-1 subtype B reference and variant calling using the Exatype bioinformatics pipeline (Hyrax Biosciences).

HIV-1 subtypes were assigned by BLAST searches against the National Center for Biotechnology Information (NCBI) 2009 HIV subtype reference sequences supplemented with NCBI reference genomes for HIV-1 subtypes A3, A4, and A6. The subtype was established using the first 1257 bases of the HIV-1 pol gene. DRMs were defined as those with a score of ≥30 in the Stanford HIV Drug Resistance Database (version 9.6) [24]. Statistical analysis was performed using Analyze-it for Excel (version 6.15) or in R (version 4.2.2) software.

RESULTS

Demographics

Most specimens submitted for HIV-1 plasma RNA or proviral DNA GRT were from men (76.3% for RNA and 78.6% for DNA). The median age (interquartile range) of individuals who submitted specimens for DNA GRT was higher than that for RNA testing (51 [38–59] vs 40 [30–52] years, respectively; P < .001) (Supplementary Table 1). Most sequences (93.9% for RNA and 95.3% for DNA) were HIV-1 subtype B, and the most frequent non-B subtypes were AG, C, AE, BF, and BG. Subtype A6, which has been identified as a baseline predictor of virologic failure for long-acting rilpivirine and cabotegravir regimens [25, 26], was found in 0.14% of RNA and 0.07% of DNA sequences.

Over the study period, the prevalence of DRMs for any drug class varied across the 10 US HHS regions (P < .001 by Pearson χ² test for RNA and DNA sequences). The highest DRM prevalence was in the Northeast (HHS region 1) for RNA sequences (34.1%) and in the Midwest (HHS region 7) for DNA sequences ( 37.3%) (Supplementary Figure 1).

DRM Trends in RNA and DNA GRT

The prevalence of resistance to any NRTI, NNRTI, PI, or INSTI declined significantly from 2018 to May 2024 for both plasma RNA and proviral DNA sequences (Figure 1). DRM prevalence was higher for DNA than for RNA at the clinically reportable MV threshold of 10% for DNA NGS. However, when the threshold was increased to 20% to approximate the sensitivity of plasma RNA Sanger sequencing for detecting MVs, the trendlines for RNA and DNA resistance were correlated (RNA vs DNA 10%, Pearson r = 0.92 [95% confidence interval, .52–.99]; RNA vs DNA 20%, Pearson r = 0.89 [.43–.98]) (Figure 1).

Figure 1.

Prevalence of resistance in human immunodeficiency virus (HIV) 1 plasma RNA and proviral DNA specimens from 2018 to 2024, including nucleoside reverse-transcriptase (RT) inhibitor, nonnucleoside RT inhibitor, protease inhibitor, and integrase strand transfer inhibitor drug resistance mutations (DRMs) with a Stanford HIV Drug Resistance Database score ≥30. RNA DRMs included plasma RNA DRMs detected by Sanger sequencing in all RNA sequences with a concurrent order for RT, protease and integrase genotypic resistance testing; DNA DRMs included those at the clinical minority variant (MV) cutoff of 10% and at an MV cutoff of 20% (Cochran-Armitage test for trends, P < .001 for RNA and DNA [both MV cutoffs]).

Most ART regimens recommend the combined use of agents in 2 classes of ARVs, such as an NRTI or NRTIs combined with an NNRTI or an INSTI [17 , 18 ]. The prevalence of dual-class resistance for the NRTIs combined with either NNRTIs or INSTIs declined from 8.7% to 4.7% for RNA and 13.1% to 8.5% for DNA sequences between 2018 and May 2024 Cochran-Armitage test for trends, P < .001) (Figure 2A). Dual- and triple-class resistance was more prevalent in older adults (aged 60–90 years) than in younger age groups (Figure 2B). The highest prevalence was noted for the dual NRTI + NNRTI category, reaching 14.1% in DNA sequences for the 60–90-year age group, compared with only 9.6% for the 40–59-year and 3.8% for the 18–39-year age groups. Single-class NRTI or NNRTI and dual-class NRTI and NNRTI resistance in both RNA and DNA sequences declined throughout the study period (Supplementary Figures 2–4).

Figure 2.

Multiclass resistance by year and by age group. A, Dual-class resistance—nucleoside reverse-transcriptase (RT) inhibitor (NRTI) + nonnucleoside RT inhibitor (NNRTI) or NRTI + integrase strand transfer inhibitor (INSTI) resistance—for RNA and DNA sequences. DNA drug resistance mutations (DRMs) with a Stanford HIV Drug Resistance Database [HIVDB] score ≥30 were analyzed at the 10% clinical minority variant (MV) cutoff. RNA DRMs included all RNA sequences with a concurrent order for RT, protease, and integrase genotypic resistance testing (Cochran-Armitage test for trends, P < .001 for RNA and DNA). B, Dual- and triple-class resistance by patient age group for RNA and DNA specimens. DRMs with a Stanford HIVDB score ≥30 were tabulated for each category. A 10% MV cutoff was used for DNA sequences (Cochran-Mantel-Haenszel test for resistance category differences by age group, P < .004 for RNA and P < .001 for DNA; adjusted χ² by age group, P < .001 for both RNA and DNA).

NRTI DRMs were more prevalent in DNA than in RNA sequences (Supplementary Figure 2). NNRTI DRMs were more prevalent in DNA than in RNA sequences from individuals with concurrent sequences for both RT/PR and integrase testing, but the prevalence was highest in the full set of RNA sequences that included those without a linked integrase sequence (Supplementary Figure 3). Dual-class NRTI + NNRTI resistance was highest in DNA sequences (Supplementary Figure 4). Although major PI DRM prevalence has declined from 5.3% to 2.1% in DNA sequences and remained <3% in RNA sequences (Supplementary Figure 5), triple-class PI resistance was 3.9% in older adults for DNA sequences, compared to <0.5% for RNA sequences (Figure 2B).

Although DNA reads were prefiltered for APOBEC-context hypermutation [27, 28], we investigated the impact of APOBEC mutations on DNA DRMs. The prevalence of APOBEC mutations at most APOBEC-associated DRM positions was <2%, with the exceptions of the integrase mutation G163R, RT mutation D67N, and PR mutation M46I (Figure 3) [29]. Overall, 19.1% of DNA sequences at the 10% MV cutoff harbored an APOBEC-context mutation, compared with 7.9% of RNA sequences (Pearson χ² test, P < .001) and only 12.3% of DNA sequences at the 20% MV cutoff (P < .001 for comparison with RNA sequences).

Figure 3.

APOBEC-associated drug resistance mutations (DRMs) at 16 of 18 resistance-associated codons in the human immunodeficiency virus (HIV) 1 integrase, reverse-transcriptase (RT), and protease genes designated as APOBEC-context mutations in the Stanford HIV Drug Resistance Database [29]. DRMs include mutations in plasma RNA sequences and mutations in proviral DNA sequences with a minority variant (MV) cutoff of 10% or 20%. Only a single instance of APOBEC-associated protease mutation G48S was found in RNA and proviral DNA sequences. The D232N integrase APOBEC-associated mutation could not be evaluated for proviral DNA sequences, as it is identical to the reference sequence. D232N is found in 0.25% of RNA integrase sequences. Abbreviations: INSTI, integrase strand transfer inhibitor; NNRTI, nonnucleoside RT inhibitor; NRTI, nucleoside RT inhibitor; PI, protease inhibitor.

Signature DRMs

The prevalence of the NRTI signature mutations, K65R/N and M184V/I, associated with resistance to the commonly used NRTIs abacavir, tenofovir, lamivudine, and emtricitabine, decreased from 2018 to 2024 in both RNA and DNA sequences (Table 1). By 2024, <1% of either RNA or DNA sequences harbored the K65R/N mutation, while M184V/I was found in 8.6% of RNA and 12.5% of DNA sequences at the MV cutoff of 10% (Table 1). When the MV cutoff for M184V/I was raised to 20% to approximate the sensitivity of Sanger sequencing, the annual prevalence of M184V/I in DNA was much closer to that in plasma RNA (9.0% vs 8.6%, respectively, in 2024) (Table 1). The canonical thymidine analogue DRM T215F/Y, prevalent with the use of first-generation NRTIs such as zidovudine and stavudine, was uncommon in RNA (1.5% in 2018 and 0.9% in 2024). T215F/Y was more prevalent in DNA sequences, but its prevalence was also declining (8.9% in 2018 and 4.5% in 2024).

Table 1.

Annual Prevalence of Nucleoside Reverse-Transcriptase Inhibitor Signature Drug Resistance Mutations

Year	DRM Prevalence in RNA Sequences, %a			DRM Prevalence in DNA Sequences, %
	K65R/N	M184V/I	K65R/N (MV cutoff, 10%b)	M184V/I (MV cutoff, 10%b)	M184V/I (MV cutoff, 20%b)
2018	1.5	11.9	1.0	19.0	13.8
2019	1.3	11.0	1.5	18.2	13.6
2020	1.1	9.7	0.9	15.3	11.5
2021	1.0	8.9	1.1	15.1	11.7
2022	0.8	8.4	0.8	13.1	9.8
2023	0.8	8.8	0.7	12.3	9.1
2024	0.7	8.6	0.8	12.5	9.0

Abbreviations: DRM, drug resistance mutation; MV, minority variant.

^aDRMs in plasma RNA sequences (determined by Sanger sequencing) were tabulated for all specimens tested for nucleoside and nonnucleoside reverse-transcriptase inhibitor and protease inhibitor resistance.

^bMutation prevalence applying a MVs cutoff of 10% or 20% to the next-generation sequencing data.

The prevalence of signature NNRTI mutations associated with resistance to rilpivirine or doravirine remained low or decreased in both RNA and DNA sequences (Table 2). Y181C/I/V, selected by first-generation NNRTIs but which also confers resistance to rilpivirine, was the most common rilpivirine-associated DRM but was found in <3% of RNA and DNA sequences in 2024 (Table 2). Mutations at RT position 103, associated with resistance to first-generation NNRTIs, also declined in prevalence from 13.5% and 15.5% in RNA and DNA sequences, respectively, in 2018% to 11.1% and 12.3% in 2024 (Table 2). Overall, the prevalence of DRMs with a score of ≥30 for the NNRTIs doravirine and rilpivirine remained low or declined slightly, with 2024 prevalences of 2% (RNA) and 2.9% (DNA) for doravirine and 6.3% (RNA) and 10.2% (DNA) for rilpivirine (Figure 4A).

Table 2.

Annual Prevalence of Selected Nonnucleoside Reverse-Transcriptase Inhibitor Drug Resistance Mutations^a

Year	Prevalence of NNRTI DRMs, %
	L100I	K101E/P	K103N/S/T	V106A/M	E138K	Y181C/I/V	Y188F/L
Plasma RNA sequences
2018	0.69	1.8	13.5	0.32	0.85	2.9	1.3
2019	0.69	1.7	12.6	0.32	0.85	2.6	1.2
2020	0.58	1.7	11.7	0.26	0.77	2.6	1.2
2021	0.53	1.6	11.1	0.26	0.77	2.4	1.1
2022	0.52	1.6	10.9	0.21	0.76	2.2	1.1
2023	0.49	1.7	10.6	0.29	0.82	2.3	1.1
2024	0.40	1.7	11.1	0.24	0.76	2.1	1.0
Proviral DNA sequences
2018	0.90	1.9	15.5	0.60	0.67	4.3	1.5
2019	1.3	2.2	14.4	0.69	0.96	4.4	1.7
2020	0.80	1.7	13.6	0.62	0.65	3.9	1.3
2021	0.91	1.7	13.9	0.29	0.66	3.0	1.1
2022	0.88	2.0	13.3	0.45	0.93	3.3	1.1
2023	0.58	1.9	12.1	0.55	0.82	3.2	1.1
2024	0.61	2.1	12.3	0.35	1.4	2.9	1.1

Abbreviations: DRMs, drug resistance mutations; NNRTI, nonnucleoside reverse-transcriptase inhibitor.

^aDRMs associated with resistance to rilpivirine and/or doravirine with a Stanford HIV Drug Resistance Database score ≥30 and present in ≥0.5% of all reverse-transcriptase sequences. K103N/S/T, associated with resistance to first-generation NNRTIs, was also tabulated.

Figure 4.

A, Doravirine and rilpivirine drug resistance mutation (DRM) prevalence in RNA and DNA sequences. B, Cabotegravir DRM prevalence in RNA and DNA sequences. Rilpivirine, doravirine, and cabotegravir DRMs with a Stanford HIV Drug Resistance Database score ≥30 were tabulated. A 10% minority variant cutoff was used for DNA sequences. Doravirine DRMs included V106A/M, Y188F/L, G190E/Q, L234I, F227C/I/L/V, M230L, Y318F; rilpivirine DRMs: L100I, K101E/P, Y181C//F/G/I/S/V, Y188L/F, G190E/Q, F227C, and M230I/L; cabotegravir DRMs, G118R, F121C, G140R, Q148H/K/R, N155H, and R263K.

The overall prevalence of INSTI resistance declined from 7.9% in RNA sequences (2018) to 5.1%, (Cochran-Armitage test for trends, P < .001) while INSTI resistance in DNA sequences remained low and relatively unchanged throughout the study period (3.5% in 2018 and 3.3% in 2024) (Supplementary Figure 6). The prevalence of dual-class NRTI and INSTI resistance also declined (Supplementary Figure 7). The prevalence of INSTI signature DRMs Q148H/K/R and H155H/T declined to 1.2% and 1.0%, respectively, in RNA sequences and <1% in DNA sequences (Table 3). However, the prevalence of R263K, which is selected by dolutegravir [30] and confers resistance to second-generation INSTIs bictegravir and cabotegravir [31–33], has increased since 2018, from 0.5% to 1.5% in RNA sequences and from 0.3% to 0.6% in DNA sequences (Table 3). Overall, DRM prevalence for the newest INSTI cabotegravir, used in long-acting formulations, remained low or slightly declined to 3.9% (RNA) and 2.5% (DNA) in 2024 (Figure 4B). Likewise, the proportions of specimens with a major INSTI DRM for dolutegravir (G118R, Q148H/K/R, and R263K) showed little variation and remained low in RNA specimens (2.7% in 2018 and 2.8% in 2024) and DNA specimens (1.2 and 1.4%, respectively).

Table 3.

Annual Prevalence of Integrase Strand Transfer Inhibitor Drug Resistance Mutations

Year	DRM Prevalence in RNA Sequences, %a			DRM Prevalence in DNA Sequences, %
	Q148H/K/R	N155H/T	R263K	Q148H/K/R	N155H/T	R263K
2018	2.1	2.2	0.5	0.8	0.6	0.3
2019	1.8	1.6	0.6	0.9	0.9	0.3
2020	1.7	1.7	0.9	0.8	0.9	0.4
2021	1.5	1.2	1.1	0.6	0.8	0.5
2022	1.4	0.9	1.2	0.6	0.8	0.6
2023	1.6	1.1	1.0	0.5	0.6	0.6
2024	1.2	1.0	1.5	0.7	0.8	0.6

Abbreviation: DRM, drug resistance mutation.

^aDRMs in plasma RNA sequences were tabulated for all specimens tested for integrase strand transfer inhibitor resistance.

DISCUSSION

Despite advances in effective ART for HIV-1, drug resistance continues to present a challenge for achieving virologic suppression in PWH. This study describes a trend of declining HIV-1 ARV resistance in the United States since 2018, consistent with the increased usage of regimens with higher resistance barriers, improved tolerability, and more convenient dosing. Overall resistance prevalence was highly correlated for sequences submitted for plasma RNA GRT by Sanger sequencing and for archived DNA GRT by NGS when the MV cutoff was set at 20% to mimic the sensitivity of Sanger sequencing. At the lower MV cutoff of 10% used for clinical reporting, resistance rates were higher in DNA sequences, but the declining trend over the study period followed the same trajectory. It is possible that residual APOBEC-context mutations accounted for the higher prevalence of DNA DRMs at the 10% MV cutoff despite filtering for hypermutated sequence reads, as well as reads harboring stop codons, frameshifts, large deletions, or enzyme active site mutations. However, the prevalence of APOBEC-context mutations at key DRMs in DNA was low, reaching only 1.4% for M184I (10% MV cutoff), compared with 1.1% in RNA sequences. Therefore, the higher DRM rates in proviral DNA at the lower cutoff likely reflect the greater sensitivity of NGS compared with Sanger sequencing.

The prevalence of NRTI mutations K65R/N and M184V/I that have the greatest impact on tenofovir, lamivudine, and emtricitabine resistance also declined. This may reflect the increased use of combination regimens that include an NRTI and an INSTI, such as dolutegravir or bictegravir, with a higher barrier to resistance than earlier-generation NNRTI or INSTI regimens, and it may also reflect the increased use of NRTI-free regimens. The use of the 2-drug combination regimen dolutegravir and lamivudine is contraindicated in patients with a history of lamivudine or dolutegravir resistance [34]. Nevertheless, the impact of M184V/I mutations on this combination is being investigated. A recent study described virologically suppressed patients with previous lamivudine resistance who were switched to dolutegravir and lamivudine [35]. After exclusion of patients with an M184V/I mutation detected by Sanger sequencing of proviral DNA, it was found that virologic suppression was maintained at week 48, and there was no treatment-emergent resistance, as determined by proviral DNA NGS of patients who discontinued treatment due to lack of efficacy [35]. Further studies are needed to define the role of DNA GRT in stably suppressed patients being considered for a switch to an INSTI and lamivudine- or emtricitabine-containing regimen.

While rilpivirine has been approved since 2011, the prevalence of major rilpivirine DRMs— including L100I, K101E/P, E138K, Y181C/I/V, and Y188F/L—has remained low in both RNA and DNA specimens submitted for reference laboratory GRT. In contrast, in PWH who experienced virologic failure while receiving rilpivirine as their sole NNRTI, the prevalence of rilpivirine DRMs was found to be high, reaching 32.2% for E138K, 17.8% for K101E, and 17.2% for Y181C [36]. Doravirine, the newest NNRTI approved by the US Food and Drug Administration since 2018, has a mostly nonoverlapping resistance profile with rilpivirine and may be useful in an NNRTI-based regimen for individuals with rilpivirine resistance. Clinical trials are also ongoing for doravirine coadministered once daily with the investigational nucleoside RT translocation inhibitor ilslatravir [37]. In the first year of clinical use for doravirine (2018–2019), 92.5%–96.7% of specimens undergoing routine phenotypic testing for susceptibility to doravirine were found to be susceptible, using 3- to 15-fold biological cutoffs [38]. As of 2024, resistance to doravirine has remained low (2.0% for RNA and 2.9% for DNA sequences), preserving this NNRTI as a treatment option for PWH with rilpivirine resistance. Although first-generation NNRTIs (efavirenz and nevirapine) are no longer commonly used in the United States, the prevalence of K103N, the signature DRM for these NNRTIs, has continued to decline but remained above 10%. K103N is among the most commonly transmitted drug resistance NNRTI mutations and has been found in 8.6% of treatment-naive PWH in the United States from 2014–2018 and in 9.1% of European PWH [11, 39], likely reflecting its minimal effect on viral fitness [40].

INSTI-based ART is now frequently prescribed both for initial treatment and for patients with resistance to other ARV classes. In addition, a long-acting formulation of cabotegravir plus rilpivirine was approved in 2021 [17, 18]. In real-world settings, the rates of virologic failure with long-acting cabotegravir-rilpivirine were found to be low, at about 1% [41, 42]. However, a meta-analysis found that INSTI resistance emerged in 40%–70% of individuals experiencing virologic failure with long-acting cabotegravir-rilpivirine [43]. In the present study, we found that the signature mutations for first-generation INSTIs at integrase positions 148 and 155 declined in RNA and remained below 1% in DNA sequences. However, the prevalence of the R263K mutation, which is selected by dolutegravir and confers cross-resistance to bictegravir and cabotegravir [33, 44], has increased for both RNA and DNA sequences. The ROSETTA registry observational study, investigating resistance patterns in individuals in whom second-generation INSTIs failed, found that virologic failure via the Q148H/K/R pathway was more common in individuals with prior exposure to first-generation INSTIs than in those who had been exposed only to second-generation INSTIs, who tended to develop the R263K or G118R mutations [45]. The switch to more frequent use of second-generation INSTIs is consistent with this trend. However, dolutegravir resistance is considered rare in treatment-naive patients [46] and in patients receiving 2- and 3-drug regimens that include dolutegravir [47]. Increases in the prevalence of R263K may affect future INSTI treatment and PrEP options and should continue to be monitored.

The prevalence of dual-class resistance to an NRTI and an NNRTI or an INSTI has also declined, and by 2024 only 4.7% of RNA sequences and 8.5% of DNA sequences harbored DRMs for dual-category multiclass resistance. There was a significant difference by age, however, with higher levels of dual- and triple-class resistance in sequences obtained from older adults. This age-related trend likely reflects longer durations of treatment with earlier-generation ARVs in the older group. As PWH in younger age groups started therapy on contemporary regimens with higher barriers to resistance and better tolerability than first-generation NRTI, NNRTI, and PI-based regimens, future resistance rates as this cohort ages may remain lower. Integrase GRT may be ordered with or without a concurrent order for PR and RT GRT. We noted that NRTI, NNRTI, and dual-class NRTI + NNRTI resistance was higher in the superset of specimens with or without an accompanying integrase GRT. These orders may include a higher proportion of individuals receiving NNRTI regimens without an INSTI, whereas concurrent orders for INSTI testing may include individuals who have been treated with an INSTI regimen. Thus, it is reasonable to expect higher prevalences of NNRTI resistance in the former group. Similarly, INSTI resistance was more prevalent in the superset of all integrase test orders that included those without concurrent RT and PR GRT, likely reflecting the higher proportion of individuals on INSTI regimens in this group.

This study had several limitations. First, it was retrospective and was limited to specimens submitted by a medical provider for RNA or DNA GRT. No patient clinical histories were available, and thus we cannot rule out that the downward trend in resistance prevalence in RNA specimens could have been confounded by increases in the proportion of treatment-naive PWH undergoing GRT while fewer treatment-experienced PWH with virologic failure required GRT. However, the downward trend was similar for proviral DNA GRT, which is primarily used in virologically suppressed, treatment-experienced PWH. Had the declines in plasma RNA resistance prevalence been driven primarily by increased testing of treatment-naive patients, these trends would not have been expected to mirror the proviral DNA testing trends. We also note that in the United States and other high-income countries, established practice guidelines [17, 18] have long-recommended genotypic RTI and PI GRT for all treatment-naive PWH at diagnosis and before initiation of ART; therefore, guideline-driven changes in RTI and PI GRT practices are less likely. Resistance testing guidelines have been updated in recent years to recommend INSTI GRT following HIV infection while using cabotegravir PrEP or for PWH who acquired HIV after discontinuing cabotegravir PrEP [18]. However, the number of cases of HIV infection in patients receiving cabotegravir PrEP has been very low [41, 48].

The use of proviral DNA sequencing for HIV GRT also has a number of limitations. DNA sequencing may underestimate the presence of DRMs due to the dilution of resistant species in the DNA reservoir of ancestral viral populations [49–51]. DNA GRT has also been found in multiple studies to be less sensitive than historical plasma RNA GRT [50]. Consequently, DNA GRT may have a lower negative predictive value than RNA GRT [51]. It may also overestimate some resistance mutations due to the presence of APOBEC3-induced defective viruses [49–51]. In the current study, we used a filtering algorithm to remove hypermutated reads that likely resulted from APOBEC3 editing, as well as reads with stop codons, frameshifts, and large deletions that likely originated from defective viruses. Although this strategy may not remove all defective viral sequences, we noted that, after filtering, the prevalence of mutations at 16 APOBEC-context positions was <2% in most cases and was highest at positions that are not considered to be major DRMs for contemporary ARVs (eg, RT D67N, with prevalences of 4.9% and 4.0% at 10% and 20% MV cutoffs). Despite these limitations, the trends in DNA resistance prevalence in this study mirrored those for RNA, with higher prevalence at the 10% reportable MV cutoff and near-identical prevalence at the higher 20% cutoff, which mimics the sensitivity of Sanger sequencing. In a comparative analysis of plasma RNA GRT versus DNA GRT performed by NGS in a cohort of 190 patients, it was found that DNA GRT detected more DRMs overall at lower sensitivity thresholds, which is also consistent with our findings [52].

To our knowledge, this is the first study to compare aggregate population trends of DNA resistance prevalence with that from routine RNA GRT. The close correlation of the declining trend of DRM prevalence in both RNA and DNA suggests that, despite the limitations of DNA GRT, shifts in aggregate population RNA DRM prevalence are also mirrored in the archived DNA compartment. Further studies are needed to explore whether this aggregate relationship holds for all DRM classes or is limited to only certain DRMs or DRM classes. Overall, DNA GRT constituted approximately 20% of the total orders for RT, PR, and integrase GRT. Increased DNA GRT ordering in recent years (data not shown) is likely driven by increasing numbers of virologically suppressed patients seeking regimen changes to long-acting injectables.

As another possible limitation, regional differences in resistance prevalence between US HHS regions were noted, but they were not large. A number of factors could account for these differences, such as differences in age distributions or adherence patterns; however, a further investigation was beyond the scope of this study.

In summary, we have shown that the prevalence of ARV resistance in a large US reference laboratory database declined from 2018 to 2024 in conjunction with the availability of novel ART regimens, including long-acting regimens and formulations with improved tolerability that promote better adherence. These novel ART regimens are noted to have enhanced effectiveness and activity against viruses resistant to earlier-generation ARVs. Prevalence of INSTI DRMs was low, supporting US guidelines for limiting INSTI GRT to cases of treatment failure, suspected transmitted resistance, or infection during INSTI-based PrEP. Resistance to ≥2 classes of ARV has also declined, continuing trends described in previous work from 2003–2012 [21], 2006–2017 [20], and 2012–2018 [19]. We have augmented our analysis by presenting resistance trends in both RNA and proviral DNA to show that the DNA resistance trends and DRM prevalences largely mirrored those for RNA. Continued advances in ART efficacy, durability, and tolerability may lead to increased rates of virologic suppression and further reduce the incidence of archived resistance mutations in proviral DNA. Large, real-world resistance test result databases are an important tool to provide insights into DRM prevalence that may affect contemporary regimens and those currently in development, as well as treatment guidelines.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.