Comparison of treatment‐emergent resistance‐associated mutations and discontinuation due to adverse events among integrase strand transfer inhibitor‐based single‐tablet regimens and cabotegravir + rilpivirine for the treatment of virologically suppressed people with HIV: A systematic literature review and network meta‐analysis

Abstract

Objective

This study evaluated rates of treatment‐emergent resistance‐associated mutations (TE‐RAMs) and discontinuation due to adverse events (DC‐AEs) across integrase strand transfer inhibitor (INSTI)‐based single‐tablet regimens and injectable cabotegravir + rilpivirine (CAB + RPV) in virologically suppressed people with HIV.

Methods

A systematic literature review was conducted for phase 2–4 randomized controlled trials with ≥48 weeks of follow‐up involving virologically suppressed people with HIV aged ≥12 years and published January 2003–March 2024. A random‐effects network meta‐analysis estimated comparative rates of TE‐RAMs and DC‐AEs among regimens at 48 weeks. Risk of bias and strength of evidence were assessed using Cochrane RoB and CINeMA, respectively.

Results

Fourteen (7509 participants) and nine (4656 participants) studies were included in the TE‐RAMs and DC‐AEs analyses, respectively. No significant differences in rates of TE‐RAMs were observed; risk ratios (RRs) for TE‐RAMs for bictegravir/emtricitabine/tenofovir alafenamide (B/F/TAF), dolutegravir/abacavir/lamivudine (DTG/ABC/3TC) and CAB + RPV every 4 weeks (Q4W) versus CAB + RPV every 8 weeks (Q8W) were 0.22 (95% CI, 0.02–2.04), 0.22 (95% CI, 0.00–19.85) and 0.40 (95% CI, 0.14–1.09). Compared with CAB + RPV Q4W and Q8W, DC‐AEs were significantly lower with B/F/TAF (RR, 0.15 [95% CI, 0.03–0.75] and RR, 0.16 [95% CI, 0.04–0.67], respectively) and DTG/ABC/3TC (RR, 0.05 [95% CI, 0.01–0.48] and RR, 0.05 [95% CI, 0.01–0.46], respectively).

Conclusions

In virologically suppressed people with HIV, switching to CAB + RPV Q8W yielded a non‐significant increased risk of TE‐RAMs compared with INSTI‐based 2‐ and 3‐drug regimens and CAB + RPV Q4W. Both CAB + RPV Q4W and Q8W had significantly higher risks of DC‐AEs than B/F/TAF and DTG/ABC/3TC. Findings highlight the importance of considering both resistance and tolerability when switching regimens.

INTRODUCTION

HIV treatment guidelines recommend integrase strand transfer inhibitor (INSTI)‐based single‐tablet regimens (STRs) as an option for first‐line therapy for most people with HIV [1]. These recommended STRs contain the INSTIs bictegravir (BIC) and dolutegravir (DTG), which both have a high barrier to resistance [1]. STRs offer the convenience of a once‐daily combination tablet, which may help to improve treatment adherence [2, 3]. In people with HIV who are adherent, STRs have been shown to be effective in maintaining durable virologic suppression, thereby preventing the emergence of resistance [2]. With the removal of dolutegravir/abacavir/lamivudine (DTG/ABC/3TC) from the US Department of Health and Human Services (HHS) list of recommended initial regimens due to the requirement for HLA‐B*5701 testing and concerns about an increased risk of cardiovascular events, bictegravir/emtricitabine/tenofovir alafenamide (B/F/TAF) remains the INSTI‐based STR with the fewest restrictions [1].

The introduction of injectable cabotegravir + rilpivirine (CAB + RPV) eliminates the need for daily oral therapy by requiring two intramuscular injections either once every 4 weeks (Q4W) or once every 8 weeks (Q8W) [1]. However, treatment‐emergent resistance‐associated mutations (TE‐RAMs), including dual‐class resistance, have been observed in people with HIV who are adherent to the injection schedule for CAB + RPV [3, 4, 5, 6, 7, 8, 9]. The development of TE‐RAMs poses significant challenges for HIV control and management using guideline‐preferred INSTI‐based regimens, as these mutations are transmissible, reduce treatment efficacy and limit future therapeutic options [10, 11, 12, 13]. Furthermore, when switching antiretroviral therapy (ART), tolerability could also be a concern for oral and injectable regimens, including adverse events unique to injectable ART (e.g., injection site reactions) [3, 5, 14, 15, 16].

An analysis of the comparative effects of different ART regimens on TE‐RAMs and discontinuation due to adverse events (DC‐AEs) has not been previously conducted. In people who are otherwise virologically controlled and tolerating current ART, it is important to understand the risk of HIV drug resistance and adverse effects across ART regimens to optimize treatment success and sustain high standards of care for people with HIV. This study aimed to conduct a network meta‐analysis (NMA) of randomized controlled trials (RCTs) to evaluate the comparative effects of different ART regimens on TE‐RAMs and DC‐AEs in virologically suppressed (VS) people with HIV.

METHODS

We conducted this systematic literature review (SLR) and NMA in accordance with the Cochrane Collaboration guidelines for systematic reviews of interventions [17] and reported results following the Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA) guidelines [18]. The study protocol was registered in PROSPERO (CRD42024571679).

Search strategy and selection criteria

We searched PubMed, Embase, Cochrane CENTRAL and EBSCO Open Dissertations for studies published from January 2003 to March 2024. An SLR was conducted for phase 2, 3 and 4 RCTs that investigated switching to any oral INSTI‐based STR (i.e., including BIC, DTG, or elvitegravir [EVG]) or injectable CAB + RPV in VS people with HIV aged ≥12 years with ≥48 weeks of follow‐up in both arms. VS was defined per study design and was routinely HIV‐1 RNA <50 copies/mL. For details on the search strategies, see Table S1.

For inclusion, the study publications must have provided sufficient data on resistance outcomes between different regimens. Arms comprised of multi‐tablet regimens were included only if the intervention arm in the study was an STR. For studies with multiple regimens in the comparison arms, the regimen with the most participants was used. Studies were excluded if the ART was investigational; participants were heavily treatment experienced based on evidence of multi‐drug resistant HIV or treatment with fostemsavir, lenacapavir, ibalizumab and/or enfuvirtide; or follow‐up time was <48 weeks.

Outcomes

The goal of this SLR and NMA was to quantify and compare the rates of TE‐RAMs and treatment DC‐AEs across INSTI‐based STRs and CAB + RPV in VS people with HIV.

Study selection and data extraction

Covidence was used to screen and select studies, and two reviewers (IR and HWS) independently screened titles, abstracts and full texts as per the predefined eligibility criteria; conflicts were resolved by a third reviewer (CW). The data extracted from studies included study characteristics (e.g., author, year of publication, study design and study duration), participant characteristics (e.g., setting, number of participants and sociodemographic characteristics), intervention and comparator characteristics (e.g., description of intervention and comparators) and outcomes of interest (e.g., rates of TE‐RAMs and DC‐AEs).

Evidence grading

We assessed the certainty of evidence using the Confidence in Network Meta‐Analysis (CINeMA) online platform. Judgements were summarized into four levels of confidence for each relative treatment effect according to the standard Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework [19]: high (the true effect closely matches the estimated effect), moderate (the estimate is likely accurate but may differ substantially), low (limited confidence with potential substantial difference from the estimate) and very low (very little confidence in the estimate, with the true effect likely being substantially different).

Quality assessment

Two reviewers (IR and HWS) independently evaluated the risk of bias (RoB) in randomized trials using the Cochrane RoB 2 tool. Disagreements were resolved by consulting a third reviewer (CW). The Cochrane RoB 2 tool assesses bias across five domains: (1) the randomization process, (2) deviations from the intended intervention, (3) missing outcome data, (4) measurement of the outcome and (5) selection of the reported result. The RoB was assessed at the study level. Each domain comprised a series of ‘signalling questions’. Judgements for each study were categorized as ‘Low’, ‘High’ or ‘Some concerns’ based on the responses to these questions [20].

Data analysis

Data analysis was conducted using ‘mvmeta’ and ‘network’ packages in Stata v17 (College Station, TX). TE‐RAMs and DC‐AEs variables were analyzed using risk ratios (RRs) with 95% CIs. Heterogeneity was evaluated using the I ² test, with high heterogeneity defined as an I ² test score >50%. A random‐effects NMA was performed using a frequentist approach to estimate comparative rates of TE‐RAMs and DC‐AEs among treatment regimens; network geometry was constructed to provide comparative evidence among the treatments in the included studies. Consistency was evaluated using the consistency‐inconsistency model [21].

We estimated the probability of which regimens are best by using the surface under the cumulative ranking curve (SUCRA) based on the rates of TE‐RAMs and DC‐AEs. SUCRA scores signal the probability a treatment has of being among the best options in the network with higher scores representing a better ranking [22, 23].

We also conducted transitivity assessments to explore the distribution of effect modifiers (i.e., baseline CD4+ counts and treatment switching) that might affect the outcomes of interest across treatment comparisons [17, 24]. Publication bias was evaluated for small‐study effects using a comparison adjusted funnel plot and Egger's test [25, 26]. A p value of <0.05 was considered statistically significant. All analyses were conducted using Stata v17. We conducted a sensitivity analysis by excluding studies within the lowest 25th percentile of sample size [27]. This approach allowed us to evaluate whether the main findings of the NMA were disproportionately influenced by these smaller trials.

RESULTS

Study selection

A total of 2555 studies were identified from the literature search. Of the identified studies, 1325 were screened and 320 were assessed for eligibility. Fourteen studies passed all selection criteria and were included in the NMA (Figure 1, Tables 1 and Table S2).

FIGURE 1.

PRISMA flow chart. ^aRegisters refers to clinical trial registries (e.g., ClinicalTrials.gov). ^bCovidence automation tool. INSTI, integrase strand transfer inhibitor; PRISMA, Preferred Reporting Items for Systematic reviews and Meta‐Analyses; RCT, randomized controlled trial; STR, single‐tablet regimen.

TABLE 1.

Characteristics of studies included in the analysis.

First author, year	Study sites	Study design	Study population	Treatment arms	CD4+ Count (cells/mm3) at baseline	Number of participants with TE‐RAMs	Number of TE‐RAMs	Sample size	Number of DC‐AEs	Sample size
Daar 2018 [28]	121 outpatient centers in 10 countries (Australia, Belgium, Canada, the Dominican Republic, France, Germany, Italy, Spain, the United Kingdom and the United States)	Randomized, open‐label, multicenter, active‐controlled, noninferiority phase 3 trial	Virologically suppressed adults	bPI + 2 NRTIs	626 (437–821)	1	1	287	1	287
				B/F/TAF	617 (469–809)	0	0	290	2	290
Molina 2018 [29]	96 outpatient centers in 9 countries (Australia, Belgium, Canada, France, Germany, Italy, Spain, the United Kingdom and the United States)	Randomized, double‐blind, multicenter, active‐controlled, noninferiority phase 3 trial	Virologically suppressed adults	DTG/ABC/3TC	661 (478–874)	0	0	281	2	281
				B/F/TAF	732 (554–936)	0	0	282	6	282
Sax 2021 [30]	94 locations (the United States)	Randomized, double‐blind, multicenter, active‐controlled, noninferiority phase 3 trial	Virologically suppressed adults	DTG + 2 NRTIs	642 (462–791)	0	0	281	6	281
				B/F/TAF	659 (486–885)	0	0	284	6	284
Arribas 2014 [31]	86 sites in Europe and North America	Randomized, open‐label, multicenter, noninferiority phase 3b trial	Virologically suppressed individuals on ART	bPI + 2 NRTIs	585 (445–770)	0	0	139	4	140
				E/C/F/TXF	564 (423–757)	0	0	290	6	293
Hodder 2018 [32]	99 locations (the United States)	Randomized, open‐label multicenter trial	Virologically suppressed women	bPI + 2 NRTIs	NR	0	0	53	1	53
				E/C/F/TXF	NR	0	0	159	1	159
Kityo 2019 [33]	58 outpatient centers in the Dominican Republic, Puerto Rico, the Russian Federation, Thailand, Uganda and the United States	Randomized, open‐label, multicenter, active‐controlled, noninferiority phase 3 trial	Virologically suppressed women	E/C/F/TXF	704 (546–878)	1	1	223	0	223
				B/F/TAF	667 (532–852)	0	0	234	0	234
Mills 2016 a [34]	168 sites in 19 countries in North America, Europe, Latin America, Asia and Australia	Randomized, open‐label, multicenter, active‐controlled noninferiority trial	Virologically suppressed adults	bPI + 2 NRTIs	662 (525–831)	0	0	199	N/A	N/A
				E/C/F/TXF	675 (520–833)	1	1	959	N/A	N/A
Pozniak 2014 a [35]	72 sites in Australia, Europe and North America	Randomized, open‐label, multicenter phase 3b trial	Virologically suppressed adults	EFV/FTC/TDF	562 (401–750)	0	0	100	N/A	N/A
				E/C/F/TXF	561 (450–722)	0	0	290	N/A	N/A
Swindells 2020 a [7]	115 locations (the United States)	Randomized, open‐label, multicenter parallel‐group trial	Treatment‐experienced, virologically suppressed adults (aged ≥18 years)	E/C/F/TXF	680 (133–2089)	1	1	75	N/A	N/A
				CAB + RPV Q4W	684 (94–1954)	1	2	308	N/A	N/A
Overton 2021 [8]	13 countries (Australia, Argentina, Canada, France, Germany, Italy, Mexico, Russia, South Africa, South Korea, Spain, Sweden and the United States)	Randomized, open‐label, multicenter, parallel‐group, noninferiority phase 3b trial	Treatment‐experienced, virologically suppressed (plasma HIV‐1 RNA <50 copies/mL) adults	CAB + RPV Q4W	688 (523–878)	2	5	523	13	523
				CAB + RPV Q8W	642 (499–827)	8	12	522	12	522
Ramgopal 2023 [9]	118 clinical centers in 14 countries (Australia, Austria, Belgium, Canada, France, Germany, Ireland, Italy, Japan, the Netherlands, Spain, Switzerland, the United Kingdom and the United States)	Randomized, open‐label, multicenter, active‐controlled, noninferiority phase 3b trial	Virologically suppressed adults	B/F/TAF	640 (459–846)	0	0	223	2	227
				CAB + RPV Q8W	649 (477–850)	3	6	447	25	454
Llibre 2023 a [36]	119 investigational centers in 17 countries (Argentina, Belgium, Brazil, Canada, China, Denmark, France, Germany, Italy, Mexico, Russia, South Africa, Spain, Sweden, Taiwan, the United Kingdom and the United States)	Randomized, open‐label, multicenter, noninferiority phase 3 trial	Virologically suppressed adults	EFV/FTC/TDF	668 (94–1954)	0	0	73	N/A	N/A
				DTG/3TC	675 (154–2089)	0	0	246	N/A	N/A
van Wyk 2020 a [37]	133 locations (the United States)	Randomized, open‐label, multicenter, noninferiority phase 3 trial	Virologic suppressed adults	E/C/F/TXF	720 (119–1810)	0	0	249	N/A	N/A
				DTG/3TC	682 (133–1904)	0	0	369	N/A	N/A
Moyle 2024 [38]	32 hospital‐based HIV clinics in 7 European countries (Belgium, France, Germany, Ireland, Italy, Spain and the United Kingdom)	Randomized, open‐label, multicenter, parallel 2‐arm trial	Virologically suppressed people with virus harbouring the Lys103Asn (K103N) mutation	bPI + 2 NRTIs	617 (284–787)	0	0	28	0	28
				DTG/RPV	510 (284–715)	0	0	95	3	95

Abbreviations: B/F/TAF, bictegravir/emtricitabine/tenofovir alafenamide; bPI, boosted protease inhibitor; CAB + RPV, cabotegravir + rilpivirine; DC‐AEs, discontinuation due to adverse events; DTG, dolutegravir; DTG/3TC, dolutegravir/lamivudine; DTG/ABC/3TC, dolutegravir/abacavir/lamivudine; DTG/RPV, dolutegravir/rilpivirine; E/C/F/TXF, elvitegravir/cobicistat/emtricitabine/(tenofovir disoproxil fumarate or tenofovir alafenamide); EFV/FTC/TDF, efavirenz/emtricitabine/tenofovir disoproxil fumarate; N/A, not applicable; NR, not reported; NRTI, nucleoside reverse transcriptase inhibitor; Q4W, every 4 weeks; Q8W, every 8 weeks; RR, risk ratio; TE‐RAM, treatment‐emergent resistance‐associated mutation.

^a Study was excluded from the analysis of the rate of DC‐AEs due to insufficient data for each regimen in the control arm.

Analysis of TE‐RAMs

All treatment regimen comparisons for the 14 RCTs included in the analysis of rates of TE‐RAMs are shown in the network map (Figure 2) [7, 8, 9, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38]. The treatment with the most participants was elvitegravir/cobicistat/emtricitabine/tenofovir disoproxil fumarate or tenofovir alafenamide (E/C/F/TXF), and the regimens compared most often were E/C/F/TXF and a boosted protease inhibitor (bPI) + 2 nucleoside reverse transcriptase inhibitors (NRTIs) [7, 31, 32, 33, 34, 35, 37].

FIGURE 2.

Network map of TE‐RAMs at 48 weeks. Node size is proportional to the number of participants across all included studies for an intervention, and line thickness is proportional to the number of studies that compared the two interventions. B/F/TAF, bictegravir/emtricitabine/tenofovir alafenamide; bPI, boosted protease inhibitor; CAB + RPV, cabotegravir + rilpivirine; DTG, dolutegravir; DTG/3TC, dolutegravir/lamivudine; DTG/ABC/3TC, dolutegravir/abacavir/lamivudine; DTG/RPV, dolutegravir/rilpivirine; E/C/F/TXF, elvitegravir/cobicistat/emtricitabine/(tenofovir disoproxil fumarate or tenofovir alafenamide); EFV/FTC/TDF, efavirenz/emtricitabine/tenofovir disoproxil fumarate; NRTI, nucleoside reverse transcriptase inhibitor; Q4W, every 4 weeks; Q8W, every 8 weeks; TE‐RAM, treatment‐emergent resistance‐associated mutation.

At 48 weeks, there were no statistically significant differences between treatment regimens for TE‐RAM RRs (Tables 2 and S3). The RRs for TE‐RAMs with B/F/TAF and DTG/ABC/3TC compared with CAB + RPV Q8W were 0.22 (95% CI, 0.02–2.04) and 0.22 (95% CI, 0.00–19.85), respectively. The RR for TE‐RAMs with CAB + RPV Q4W compared with CAB + RPV Q8W was 0.40 (95% CI, 0.14–1.09). The certainty (quality) of evidence for each comparison in TE‐RAMs was judged as very low or low, driven mostly by ‘some concerns’ and ‘major concerns’ for within‐study bias and imprecision, due to the open‐label study design for 12 of the 14 studies, few head‐to‐head comparisons and wide CIs (Table S4). Tests for inconsistency (p = 0.972) and residual heterogeneity (I ²: 0.00, p = 1.00) were not significant.

TABLE 2.

Pooled estimates of risk of TE‐RAMs and DC‐AEs at 48 weeks for B/F/TAF, CAB + RPV and DTG/ABC/3TC.

		Pooled estimates (RR [95% CI]) a
Comparison regimen		TE‐RAMs	DC‐AEs
B/F/TAF versus	EFV/FTC/TDF	0.13 (0.00–5.73)	–
	bPI + 2 NRTIs	0.20 (0.02–1.97)	1.46 (0.18–11.49)
	CAB + RPV Q8W	0.22 (0.02–2.04)	0.16 (0.04–0.67)
	E/C/F/TXF	0.34 (0.04–2.68)	2.16 (0.24–19.42)
	DTG/3TC	0.46 (0.01–20.78)	–
	CAB + RPV Q4W	0.55 (0.06–5.33)	0.15 (0.03–0.75)
	DTG/RPV	0.67 (0.01–61.05)	0.69 (0.02–24.91)
	DTG + 2 NRTIs	0.99 (0.02–49.69)	0.99 (0.32–3.03)
	DTG/ABC/3TC	1.00 (0.02–50.04)	2.99 (0.61–14.68)
CAB + RPV Q4W versus	EFV/FTC/TDF	0.23 (0.01–10.30)	–
	bPI + 2 NRTIs	0.37 (0.03–4.86)	9.85 (0.71–136.53)
	CAB + RPV Q8W	0.40 (0.14–1.09)	1.08 (0.50–2.35)
	E/C/F/TXF	0.62 (0.08–4.75)	14.61 (0.95–224.67)
	DTG/3TC	0.84 (0.02–37.35)	–
	DTG/RPV	1.23 (0.01–130.99)	4.66 (0.09–239.35)
	DTG + 2 NRTIs	1.81 (0.02–168.10)	6.69 (0.93–48.22)
	DTG/ABC/3TC	1.82 (0.02–169.29)	20.20 (2.07–196.86)
	B/F/TAF	1.83 (0.19–17.86)	6.76 (1.33–34.42)
CAB + RPV Q8W versus	EFV/FTC/TDF	0.59 (0.01–27.44)	–
	bPI + 2 NRTIs	0.93 (0.07–12.83)	9.11 (0.74–112.34)
	E/C/F/TXF	1.55 (0.18–13.25)	13.51 (0.98–185.72)
	DTG/3TC	2.12 (0.05–99.48)	–
	CAB + RPV Q4W	2.52 (0.92–6.90)	0.92 (0.43–2.01)
	DTG/RPV	3.08 (0.03–338.48)	4.31 (0.09–204.95)
	DTG + 2 NRTIs	4.55 (0.05–414.62)	6.18 (1.00–38.06)
	DTG/ABC/3TC	4.59 (0.05–417.54)	18.68 (2.20–158.91)
	B/F/TAF	4.60 (0.49–43.19)	6.25 (1.49–26.15)
DTG/ABC/3TC versus	EFV/FTC/TDF	0.13 (0.00–30.09)	–
	bPI + 2 NRTIs	0.20 (0.00–18.81)	0.49 (0.04–6.61)
	CAB + RPV Q8W	0.22 (0.00–19.85)	0.05 (0.01–0.46)
	E/C/F/TXF	0.34 (0.00–28.46)	0.72 (0.05–10.89)
	DTG/3TC	0.46 (0.00–109.05)	–
	CAB + RPV Q4W	0.55 (0.01–50.90)	0.05 (0.01–0.48)
	DTG/RPV	0.67 (0.00–264.53)	0.23 (0.00–11.67)
	DTG + 2 NRTIs	0.99 (0.00–252.57)	0.33 (0.05–2.32)
	B/F/TAF	1.00 (0.02–50.40)	0.33 (0.07–1.64)

Abbreviations: B/F/TAF, bictegravir/emtricitabine/tenofovir alafenamide; bPI, boosted protease inhibitor; CAB + RPV, cabotegravir + rilpivirine; DC‐AEs, discontinuation due to adverse events; DTG, dolutegravir; DTG/3TC, dolutegravir/lamivudine; DTG/ABC/3TC, dolutegravir/abacavir/lamivudine; DTG/RPV, dolutegravir/rilpivirine; E/C/F/TXF, elvitegravir/cobicistat/emtricitabine/(tenofovir disoproxil fumarate or tenofovir alafenamide); EFV/FTC/TDF, efavirenz/emtricitabine/tenofovir disoproxil fumarate; NRTI, nucleoside reverse transcriptase inhibitor; Q4W, every 4 weeks; Q8W, every 8 weeks; RR, risk ratio; TE‐RAM, treatment‐emergent resistance‐associated mutation.

^a RR (95% CI) values are for first column: second column. Bolding signifies statistically significant RRs.

According to SUCRA scores, B/F/TAF ranked the lowest for the probability of developing TE‐RAMs, and efavirenz/emtricitabine/tenofovir disoproxil fumarate ranked the highest. CAB + RPV Q4W ranked similar to DTG/3TC, DTG/RPV and DTG‐based 3‐drug regimens; however, CAB + RPV Q8W showed a higher probability of developing TE‐RAMs than all INSTI‐ and bPI‐based regimens (Table 3 and Figure S1).

TABLE 3.

SUCRA rankings for TE‐RAMs and DC‐AEs at 48 weeks.

Treatment regimen a	SUCRA score (%) b
	TE‐RAMs	DC‐AEs
B/F/TAF	71.6	53.7
DTG/ABC/3TC	64.2	83.3
DTG + 2 NRTIs	63.8	53.9
CAB + RPV Q4W	58.8	10.9
DTG/RPV	57.2	45.2
DTG/3TC	52.0	–
E/C/F/TXF	44.4	76.8
bPI + 2 NRTIs	31.0	63.1
CAB + RPV Q8W	30.9	13.1
EFV/FTC/TDF	26.2	–

Abbreviations: B/F/TAF, bictegravir/emtricitabine/tenofovir alafenamide; bPI, boosted protease inhibitor; CAB + RPV, cabotegravir + rilpivirine; DC‐AEs, discontinuation due to adverse events; D/C/F/TAF, darunavir/cobicistat/emtricitabine/tenofovir alafenamide; DTG, dolutegravir; DTG/3TC, dolutegravir/lamivudine; DTG/ABC/3TC, dolutegravir/abacavir/lamivudine; DTG/RPV, dolutegravir/rilpivirine; E/C/F/TXF, elvitegravir/cobicistat/emtricitabine/(tenofovir disoproxil fumarate or tenofovir alafenamide); EFV/FTC/TDF, efavirenz/emtricitabine/tenofovir disoproxil fumarate; NRTI, nucleoside reverse transcriptase inhibitor; Q4W, every 4 weeks; Q8W, every 8 weeks; SUCRA, surface under the cumulative ranking curve; TE‐RAM, treatment‐emergent resistance‐associated mutation.

^a Regimens are listed in order of decreasing SUCRA ranking for TE‐RAMs analysis.

^b SUCRA scores signal the probability a treatment has of being among the best options in the network, with higher scores representing better ranking.

RoB analysis

All 14 studies (7509 participants) were used in the analysis of rates of TE‐RAMs. Twelve out of the 14 studies had ‘some concerns’ in the overall RoB domain, which was largely attributable to an open‐label study design (Figure S2).

Transitivity analysis

The transitivity for TE‐RAMs was evaluated based on a subjective assessment of various effect modifiers, such as baseline CD4+ counts and treatment switching. These effect modifiers were consistent across studies, indicating that the transitivity assumption was met, and providing confidence in the comparability of treatment effects across studies.

Publication bias

The publication bias for TE‐RAMs was evaluated using both graphical and statistical techniques. A visual assessment showed a symmetrical funnel plot, confirming the absence of publication bias. Additionally, Egger's regression test did not yield a statistically significant intercept (p = 0.940), suggesting no evidence of small‐study effects or publication bias across the studies included in the network (Figure S3).

Sensitivity analysis

Three studies were excluded from the sensitivity analysis of rates of TE‐RAMs based on the lowest 25th percentile of sample size, while 11 RCTs were included in the analysis and are shown in the network map (Figure S4) [7, 8, 9, 28, 29, 30, 31, 33, 34, 35, 37].

Results from the sensitivity analysis were consistent with the full analysis, with no statistically significant differences between treatment regimens for TE‐RAM RRs (Table S5). Tests for inconsistency were not significant (p = 0.918), indicating that direct and indirect evidence in this NMA are in agreement and support the reliability of the NMA findings. SUCRA scores were also consistent with the full analysis (Table S6).

The sensitivity analysis results demonstrate that the main findings of the NMA were not disproportionately influenced by smaller trials. The treatment effect estimates for TE‐RAMs remained consistent, suggesting that the NMA results are not significantly affected by small‐study effects, thereby enhancing the reliability and validity of our findings.

Analysis of DC‐AEs

All treatment regimen comparisons for the nine RCTs included in the analysis of rates of DC‐AEs are shown in the network map (Figure 3) [8, 9, 28, 29, 30, 31, 32, 33, 38]. The treatment with the most participants was B/F/TAF, and the regimens compared most often were E/C/F/TXF and a bPI + 2 NRTIs [9, 28, 29, 30, 31, 32, 33].

FIGURE 3.

Network map of DC‐AEs at 48 weeks. Node size is proportional to the number of participants across all included studies for an intervention, and line thickness is proportional to the number of studies that compared the two interventions. B/F/TAF, bictegravir/emtricitabine/tenofovir alafenamide; bPI, boosted protease inhibitor; CAB + RPV, cabotegravir + rilpivirine; DC‐AEs, discontinuation due to adverse events; DTG, dolutegravir; DTG/ABC/3TC, dolutegravir/abacavir/lamivudine; DTG/RPV, dolutegravir/rilpivirine; E/C/F/TXF, elvitegravir/cobicistat/emtricitabine/(tenofovir disoproxil fumarate or tenofovir alafenamide); NRTI, nucleoside reverse transcriptase inhibitor; Q4W, every 4 weeks; Q8W, every 8 weeks.

The risk of DC‐AEs at 48 weeks was significantly lower with B/F/TAF and DTG/ABC/3TC than with CAB + RPV Q4W and CAB + RPV Q8W (Tables 2 and S7). The risk was 85% and 84% lower with B/F/TAF compared with CAB + RPV Q4W (RR, 0.15 [95% CI, 0.03–0.75]) and CAB + RPV Q8W (RR, 0.16 [95% CI, 0.04–0.67]), respectively. The risk was 95% lower with DTG/ABC/3TC compared with both CAB + RPV Q4W (RR, 0.05 [95% CI, 0.01–0.48]) and CAB + RPV Q8W (RR, 0.05 [95% CI, 0.01–0.46]). Similar to TE‐RAMs, the certainty (quality) of evidence for each comparison in DC‐AEs was rated as very low or low, driven mostly by ‘some concerns’ and ‘major concerns’ for within‐study bias and imprecision, a consequence of the open‐label study design for seven of the nine included studies, few head‐to‐head comparisons and wide CIs (Table S8). Tests for inconsistency (p = 0.621) and residual heterogeneity (I ²: 0.17, p = 0.288) were not significant.

According to SUCRA scores, CAB + RPV Q4W and CAB + RPV Q8W ranked the highest for the probability of experiencing DC‐AEs (Table 3 and Figure S5).

RoB analysis

A subset of nine studies (4656 participants) was used in the analysis of rates of DC‐AEs. Similar to the TE‐RAMs assessment, seven out of the nine studies had ‘some concerns’ in the overall RoB domain, which was largely attributable to an open‐label study design. An open‐label study contributes to RoB concerns, as participants, caregivers and intervention providers were aware of the assigned interventions throughout the trial (Figure S6).

Transitivity analysis

The transitivity for DC‐AEs was evaluated based on a subjective assessment of various effect modifiers, such as baseline CD4+ counts and treatment switching. These effect modifiers were consistent across studies, indicating that the transitivity assumption was met and providing confidence in the comparability of treatment effects across studies.

Publication bias

The publication bias for DC‐AEs was evaluated using both graphical and statistical techniques. A visual assessment showed a symmetrical funnel plot, confirming the absence of publication bias. Additionally, Egger's regression test did not yield a statistically significant intercept (p = 0.882), suggesting no evidence of small‐study effects or publication bias across the studies included in the network (Figure S7).

Sensitivity analysis

Two studies were excluded from the sensitivity analysis of rates of DC‐AEs based on the lowest 25th percentile of sample size, while seven RCTs were included in the analysis, as shown in the network map (Figure S8) [8, 9, 28, 29, 30, 31, 33].

Results from the sensitivity analysis were consistent with the full analysis. The risk of DC‐AEs was significantly lower with B/F/TAF and DTG/ABC/3TC compared with CAB + RPV Q4W and CAB + RPV Q8W (Table S9). Tests for inconsistency (p = 0.661) were not significant. According to SUCRA scores, CAB + RPV Q4W and CAB + RPV Q8W ranked the highest for the probability of experiencing DC‐AEs (Table S10).

The sensitivity analysis results demonstrate that the main findings of the NMA were not disproportionately influenced by smaller trials. The treatment effect estimates for DC‐AEs remained consistent, suggesting that the NMA results are not significantly affected by small‐study effects, thereby enhancing the reliability and validity of our findings.

DISCUSSION

In this study, we performed an NMA including 14 RCTs to characterize the rate of TE‐RAMs and DC‐AEs across INSTI‐based STRs and CAB + RPV in VS people with HIV. There were no cases of TE‐RAMs with B/F/TAF and DTG/ABC/3TC in the network of RCTs. These regimens also tended to have the lowest risk of TE‐RAMs, although not statistically significant, among the CAB + RPV regimens and all the INSTI‐based STRs included in the network. Additional DTG‐based regimens (i.e., DTG/3TC, DTG/RPV) also had no cases of TE‐RAMs in the studies included, but did not rank as high as B/F/TAF and DTG/ABC/3TC. This may be attributed to smaller sample sizes, reliance on indirect estimates due to few head‐to‐head studies in the network and/or comparators ranked lower in the network (e.g., E/C/F/TXF, EFV/FTC/TDF). Of note, the DYAD study, published after the search window, directly compared B/F/TAF versus DTG/3TC in an open‐label RCT enrolling VS participants with no history of virologic failure already on B/F/TAF [39]. TE‐RAMs were observed in two participants with confirmed virologic withdrawal (one participant in each group) and in one participant without confirmed virologic withdrawal on DTG/3TC.

Regarding CAB + RPV in this analysis, the Q8W dosing regimen performed similarly to EVG‐ and efavirenz‐containing STRs with a lower barrier to resistance. Although not statistically significant, there was an observed trend in which CAB + RPV Q8W had a numerically higher risk of resistance compared with all INSTI‐based STRs and CAB + RPV Q4W. For example, the risk of TE‐RAMs with B/F/TAF, DTG/ABC/3TC and CAB + RPV Q4W may be approximately 78%, 78% and 60% lower than CAB + RPV Q8W (RR, 0.22 [95% CI, 0.02–2.04], RR, 0.22 [95% CI, 0.00–19.85] and RR, 0.40 [95% CI 0.14–1.09], respectively). Moreover, the risk of TE‐RAMs with CAB + RPV Q8W appeared higher than other two‐drug regimens (i.e., DTG/3TC and DTG/RPV) included in the network. Few direct head‐to‐head studies and overall low rates of TE‐RAMs may have limited the ability of the analysis to detect a significant difference.

While the incidence of TE‐RAMs is low among guideline‐recommended ARTs, of particular concern to clinicians is emergent INSTI resistance observed in people with HIV receiving on‐time injections with CAB + RPV. A separate meta‐analysis determined 1% of VS people with HIV (n = 98/8314) and 5% of people with HIV with viraemia (n = 37/910) who switched to CAB + RPV experienced virologic failure, which resulted in higher rates of treatment‐emergent INSTI resistance (approximately 40%–70%) than people with HIV on DTG‐based regimens. This translated to intermediate‐to‐high levels of cross‐resistance to DTG in 64% (n = 21/33) of individuals with treatment‐emergent INSTI resistance with available genotype results [40]. This rate of cross‐resistance is consistent with an analysis of 52 clinical isolates derived from cabotegravir (CAB) failures which found that 46% were not fully susceptible to BIC [41].

While this analysis focused on resistance outcomes up to 48 weeks, it is important to recognize that until there is a cure, treatment for HIV‐1 is lifelong. TE‐RAMs continue to emerge in clinical trials with CAB + RPV after the first year, but have not been observed in clinical trials with long‐term VS people with HIV receiving BIC‐ or DTG‐based regimens [42, 43, 44, 45, 46]. In the Q8W arm of the ATLAS‐2M study of CAB + RPV, an additional 6 TE‐RAMs were identified in three people with HIV with virologic failure in years 2 and 3, whereas, no additional TE‐RAMs occurred in the Q4W arm after the first year [3, 47]. Published outside the search window for this SLR, the randomized, multicenter, open‐label CARES study of VS adults switching to CAB + RPV Q8W or remaining on oral ART reported TE‐RAMs in two participants with virologic failure who were receiving CAB + RPV Q8W, while no TE‐RAMS or virologic failures were observed in those receiving DTG/3TC/TDF, EFV/3TC/TDF, or EFV/FTC/TDF through Week 48 [5].

None of the cases of TE‐RAMs with CAB + RPV in clinical trials have been attributed to delayed injection administration, suggesting that factors beyond adherence are contributing to an increased risk of TE‐RAMs with CAB + RPV [3, 5, 6, 7, 8, 9]. Potential factors that have been reported to be associated with virologic failure with CAB + RPV include HIV‐1 subtype A6/A1, pre‐existing RPV RAMs, RPV troughs in the lower quartile and a body mass index ≥30 kg/m². However, virologic failure with CAB + RPV is difficult to predict. Of people with HIV who experienced virologic failure in CAB + RPV clinical trials, 26% had no risk factors at baseline [3, 6, 7, 8, 9, 47, 48, 49]. Differences in barriers to resistance between BIC‐, DTG‐ and CAB‐based regimens may potentially be explained by individual drug factors, such as dissociation half‐lives and binding kinetics. Across several distinct studies, scintillation proximity assays were used to determine the in vitro dissociation kinetics of radioactively labelled INSTIs from HIV‐1 integrase‐DNA complexes [50, 51, 52, 53]. In these studies, BIC demonstrated the longest dissociation half‐life at 135–163 h, followed by DTG at 71–96 h and CAB at 51 h.

When switching ART in VS people with HIV, it is important for healthcare providers to consider the risk of resistance as well as safety and tolerability. In this analysis, the risk of DC‐AEs was significantly higher for CAB + RPV dosed either Q4W or Q8W than for B/F/TAF and DTG/ABC/3TC. While these discontinuations were inclusive of injection‐site reactions, participants in the injectable CAB + RPV clinical trials were cognizant of the injectable formulation prior to entering the studies, and were still more likely to discontinue CAB + RPV. No adverse event risk benefit was identified by the network for the oral two‐drug STR DTG/RPV. While not included in the DC‐AE network due to incomplete adverse event data for the mixed regimen comparator arm, DTG/3TC did not differ from B/F/TAF in the safety outcomes evaluated in the DYAD study [39]. DTG/ABC/3TC was identified as the ART with the lowest likelihood of DC‐AEs within 48 weeks. Importantly, the most recent updates to the US HHS and International Antiviral Society–USA guidelines removed DTG/ABC/3TC from the list of preferred ART for first‐line therapy, recommending people with or at high risk for cardiovascular disease to avoid abacavir or use it with caution, or switch to an active, non‐abacavir‐based regimen [1, 54]. This was due, in part, to increasing evidence of adverse cardiovascular outcomes with abacavir use, most recently highlighted by the landmark REPRIEVE study [55].

The current study has several limitations. First, a low number of events was observed for both TE‐RAMs and DC‐AEs, and as mentioned previously, it was further limited by few direct head‐to‐head studies and reliance on indirect estimates in the analysis. This likely contributed to wide confidence intervals and lack of statistical significance. Several studies were excluded due to durations less than 48 weeks (a Food and Drug Administration standard for primary outcomes) or lack of randomization after 24 weeks, thus limiting the ability to execute an analysis of long‐term data. The majority of studies were open‐label, which can be susceptible to adverse event reporting bias, particularly when some participants are switched to a new regimen and others are continued on their stable regimen. The open‐label design would not be expected to impact TE‐RAMs; yet, based on the GRADE framework, a penalty is applied resulting in a low level of certainty when interpreting the RRs. Statistical methodology for a NMA is unable to account for study arms with multiple regimens, as was the case with several clinical trials in this analysis; therefore, only the comparator regimen with the most participants was included. Additionally, the analysis of DC‐AEs was restricted to 9 out of the total 14 studies due to lack of details when comparator arms contained multiple regimens.

Within this network of RCTs in VS people with HIV, B/F/TAF and DTG/ABC/3TC had a low risk of resistance and a favourable risk profile for DC‐AEs, particularly when compared with CAB + RPV. Until there is a cure, well‐tolerated and durable antiretroviral regimens are needed for long‐term success; moreover, the occurrence of virologic failure after sustained virologic suppression can have substantial impacts on both people with HIV and healthcare providers, as well as public health implications [4]. Recognizing the impact of resistance on current and future treatment options, clinicians should include the differential risk of TE‐RAMs and tolerability in shared decision‐making discussions when switching ART in stable, suppressed individuals.

AUTHOR CONTRIBUTIONS

Conception and design: IR, NRU, CW, TD, HWS, NC. Analysis and interpretation of the data: IR, NRU, CW, TD, MR, EMS, NY, RS, ES, ARW, SN, HWS, NC. Drafting the paper or revising it critically for intellectual content: IR, NRU, CW, TD, MR, EMS, NY, RS, ES, ARW, SN, HWS, NC.

FUNDING INFORMATION

This analysis was funded by Gilead Sciences, Inc.

CONFLICT OF INTEREST STATEMENT

IR, CW, HWS and NC are employees of the University of Utah, which received a research grant from Gilead Sciences, Inc., to conduct this work; CW also received funding for travel and conference registration fees from Gilead Sciences, Inc.; and HWS is a Utah Academy of Managed Care Pharmacy legislative chair. NRU, ARW and SN are employees of and may own stocks or stock options in Gilead Sciences, Inc. TD is an adjunct faculty member of the University of Utah, which received a research grant from Gilead Sciences, Inc., to conduct this work. MR received consulting fees from AbbVie, Gilead Sciences, Inc., Merck and ViiV Healthcare; and received honoraria from Gilead Sciences, Inc., and ViiV Healthcare. NY received honoraria/consulting fees from Astellas Pharma US, Inc., and ViiV Healthcare; and received food/beverage reimbursement from Gilead Sciences, Inc., and ViiV Healthcare. RS received speakers bureau honoraria from Gilead Sciences, Inc.; received research contracts paid to her institution from Gilead Sciences, Inc., Merck and Roche; she received a project fee from IMO Core; and is a board member of the Spokane AIDS Network. EMS and ES have no conflicts to report.

Supporting information

Data S1. Supporting information.

Rashid I, Unger NR, Willis C, et al. Comparison of treatment‐emergent resistance‐associated mutations and discontinuation due to adverse events among integrase strand transfer inhibitor‐based single‐tablet regimens and cabotegravir + rilpivirine for the treatment of virologically suppressed people with HIV: A systematic literature review and network meta‐analysis. HIV Med. 2025;26(8):1184‐1198. doi: 10.1111/hiv.70050