Virologic Failure and Emergent Integrase Strand Transfer Inhibitor Drug Resistance With Long-Acting Cabotegravir for HIV Treatment: A Meta-analysis

Abstract

Background

The long-acting injectable regimen of cabotegravir plus rilpivirine (CAB/RPV) emerged as an alternative to oral standard-of-care integrase strand transfer inhibitor (INSTI)–based regimens for individuals with adherence challenges or preference for reduced dosing schedules. Although oral INSTI regimens have a high barrier to emergent resistance, less is known about the potency and durability of CAB/RPV.

Methods

We reviewed clinical trial registries, PubMed, EMBASE, and conference abstract databases to identify reports of CAB/RPV for HIV therapy. We abstracted data on virologic failure (VF) and treatment-emergent INSTI resistance at 48 weeks (range: 24–52). We used single-proportion meta-analysis to summarize outcomes in 3 populations: antiretroviral therapy (ART)–naive individuals initiating CAB/RPV following suppression on oral ART, ART-experienced individuals switched to CAB/RPV with virologic suppression, and ART-experienced individuals switched to CAB/RPV with detectable viremia. Cochrane's RoB 2.0 and ROBINS-1 tools assessed risk of bias.

Results

Thirty-three studies (N = 9224) reported VF prevalence. Nineteen studies (N = 5662) reported resistance data. VF prevalence was 1% (95% CI: 1%–3%) in induction-maintenance studies, 1% (1%–2%) in switch-suppressed studies, and 5% (3%–10%) in switch-viremic studies. INSTI resistance prevalence among successfully genotyped participants at failure was 71% (25%–95%), 61% (44%–75%), and 41% (20%–65%) respectively. Dolutegravir cross-resistance was common (64% of those with emergent resistance).

Conclusions

Although VF rates with CAB/RPV were low, INSTI resistance emerged in approximately 40%–70% of individuals experiencing VF. These rates are significantly higher than those for oral INSTI-based regimens. Both individual-level and broader resistance surveillance may be warranted in populations with expanding CAB/RPV use.

Clinical Trials Registration. PROSPERO registration CRD42024543919.

Although 48-week virological failure rates were low, emergent integrase resistance among those failing injectable cabotegravir was common (41%–71%), and cross-resistance with dolutegravir was high. Surveillance of resistance after cabotegravir failure may be critical to ensure sustainability of dolutegravir-based antiretroviral therapy.

Graphical Abstract

Graphical Abstract.

This graphical abstract is also available at Tidbit: https://tidbitapp.io/institutional-portal/clinical-infectious-diseases/tidbits/virologic-failure-and-emergent-integrase-strand-transfer-inhibitor-drug-resistance-with-long-acting-cabotegravir-for-hiv-treatment-a-meta-analysis/update

Dolutegravir (DTG)-based antiretroviral therapy (ART) is first-line therapy in much of the world, based on cost, favorable efficacy and safety profiles, and high genetic barrier to resistance [1]. However, persistent adherence challenges associated with once-daily oral regimens can compromise treatment effectiveness in some populations of people with human immunodeficiency virus (PWH) [2]. Sociostructural factors, such as inadequate housing or insurance coverage, stigma, and concurrent physical and mental health comorbidities, can compromise treatment adherence [3]. Long-acting ART preparations with less frequent dosing schedules could help some people overcome these barriers and improve virological outcomes [4].

Cabotegravir plus rilpivirine (CAB/RPV) is the first such long-acting ART regimen for human immunodeficiency (HIV) treatment. Multiple randomized controlled trials (RCTs) such as Antiretroviral Therapy as Long Acting Suppression trial (ATLAS) [5], Antiretroviral Therapy as Long Acting Suppression every 2 Months (ATLAS-2M) [6], First Long-Acting Injectable Regimen trial (FLAIR) [7], Long-Acting antireTroviral Treatment Enabling 2 trial (LATTE-2) [8], and Switch Onto Long Acting Regimen (SOLAR) [9] demonstrate the efficacy of CAB/RPV alongside high levels of patient preference for injectable regimens compared to oral therapies [10]. Recently, cohort studies have begun to demonstrate promising outcomes for individuals with persistent adherence challenges on oral therapies who transition to CAB/RPV with detectable viremia [11, 12]. Yet, less has been reported on the emergence of drug resistance with this regimen. The extended drug exposure times associated with long-acting formulations could potentially promote selection of integrase strand transfer inhibitor (INSTI)–resistant HIV strains [1]. This elevated risk has been seen in men and transgender women using long-acting injectable CAB for HIV prevention [13]. With over 25 million people estimated to be using DTG-based ART regimens as of early 2024 [14], the emergence and transmission of INSTI-resistant HIV could compromise future HIV treatment and prevention efforts by limiting treatment options [15].

To quantify the risk of emergent drug resistance with CAB/RPV use, we conducted a meta-analysis of virologic failure (VF) and treatment-emergent INSTI resistance in people receiving CAB/RPV for the treatment of HIV. We considered 3 scenarios of CAB/RPV use: (1) previously ART-naive individuals initiating CAB/RPV after achieving suppression on oral ART, (2) those switching to CAB/RPV with virologic suppression, and (3) those switching to CAB/RPV with detectable viremia. Our overarching aim was to elucidate the potential impact of increasing CAB/RPV use on the future efficacy of INSTIs as first-line therapy for HIV.

METHODS

Search Strategy and Protocol

Our study protocol was developed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Protocols (PRISMA-P) and was registered prospectively on PROSPERO (CRD42024543919).

We conducted a systematic review to identify all studies reporting VF and/or emergence of INSTI resistance approximately 48 weeks after initiating CAB/RPV for HIV treatment. A focused search strategy based on Population/Intervention/Control/Outcome/Study design (PICOS) criteria was developed and adapted to be used in Embase, PubMed, and ClinicalTrials.gov (Supplementary Methods). We screened abstract, poster, and oral presentation databases of recent conferences on HIV, the full list of which is included in the Supplementary Methods. Last, reference lists of the latest World Health Organization (WHO) reports on HIV drug resistance and treatment optimization [16, 17] were assessed. Corresponding authors were contacted for clarification of results when required.

Inclusion Criteria

We included both original observational and interventional studies assessing efficacy of long-acting injectable CAB/RPV ART in PWH of any age, background, or geographic location. Case studies and case series were excluded. A full description of inclusion and exclusion criteria is listed in the Supplementary Methods.

Data Extraction and Quality Assessment

Search results were uploaded onto the systematic review software Covidence (version 2024; Covidence, Melbourne, Australia) for duplicate removal and 2-stage screening. Titles and abstracts were screened by 2 independent reviewers (A. P. N., C. T. N.). Disagreement was resolved by discussion and consensus. Subsequent full-text screening and data extraction were carried out manually by a single reviewer (A. P. N.). Extracted data are specified in the Supplementary Methods.

Risk of bias was ascertained using the Cochrane tool for assessing risk of bias in randomized trials (RoB 2.0 [18]) and the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool. Studies at serious or critical risk of bias were excluded from the primary statistical analysis.

Outcomes

The primary outcome under assessment was emergence of major INSTI-associated drug-resistance mutations (DRMs), based on the Stanford algorithm [19] measured as a proportion of the total number of successfully genotyped participants with VF. Levels of acquired CAB resistance and DTG cross-resistance were rated as high, intermediate, low, potentially low, or susceptible based on the Stanford algorithm. The proportion of participants developing protocol-defined VF as a proportion of the number of participants randomized per treatment arm receiving at least 1 injection was assessed as the secondary outcome. All proportions were calculated as inverse-variance weighted percentages with corresponding 95% confidence intervals (CIs). When results were reported at multiple time points, we included the one nearest our primary outcome. Subgroup analyses were excluded in favor of full cohort results.

Data Synthesis and Statistical Analysis

Between-study heterogeneity was quantified using the I² statistical parameter and considered low if I² was less than 30%, moderate if 30%–50%, and high if greater than 50%.

For each outcome, we conducted 3 separate single-proportion meta-analyses, grouping studies into 3 categories: (1) induction-maintenance studies (PWH who were ART-naive prior to study recruitment and were started on CAB/RPV after achieving viral suppression on a short course of oral ART), (2) switch-suppressed studies (ART-experienced participants with virologic suppression at trial entry defined as viral load [VL] <50 copies/mL), or (3) switch-viremic studies (ART-experienced participants with detectable viremia at trial entry defined as VL >50 copies/mL). If studies included both virologically suppressed and unsuppressed participants at trial entry and did not stratify results by pretreatment VL, pooled results were included in the switch-viremic meta-analyses. Five investigatory sensitivity analyses using alternative outcome definitions were performed: first, including only studies using a standard VF definition; second, narrowing follow-up limits to 48 ± 4 weeks; third, removing follow-up limits; fourth, only including switch-viremic studies of fully viremic populations; and last, calculating treatment-emergent INSTI resistance in an intention-to-treat population.

I² was greater than 70% for efficacy analyses and 0 for resistance analyses. However, given the small sample sizes in included studies, and resulting wide CIs calculated for each point estimate, I² might not be a precise measure of between-study heterogeneity. To account for this, as well as the heterogeneity inherent to the inclusion of both real-world cohort studies and RCTs, a random-effects model was used. Given the small sample sizes and rare incidence of both VF and resistance, we applied a logit transformation with Clopper-Pearson exact CIs. Studies measuring both outcomes were included in both analyses. Meta-analysis was conducted using R Studio 2 (version 2024.04.0+735; R Foundation for Statistical Computing, Vienna, Austria).

Certainty of the Evidence

The certainty of evidence was ranked using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach across 5 different domains: risk of bias, inconsistency, indirectness, publication bias, and imprecision.

RESULTS

Literature Search

Our literature search on 9 November 2024 identified 1032 publications (Figure 1). After exclusion of duplicates, we reviewed 792 abstracts and 22 conference reports, and identified 62 reports of 59 studies assessing efficacy of long-acting CAB/RPV [5–9, 11, 12, 20–72]. Most articles excluded during full-text review were interim reports (22/38, 58%) or studies on CAB for pre-exposure prophylaxis (7/38, 18%).

Figure 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow chart for study selection. Abbreviations: CAB, cabotegravir; LATTE, Long-Acting antireTroviral Treatment Enabling; LEN, Lenacapavir; nRCT, non–randomized controlled trial; PrEP, pre-exposure prophylaxis; RCT, randomized controlled trial; RPV, rilpivirine; WHO, World Health Organization.

Fourteen studies (14/62, 23%) with 29 total participants were case reports and were excluded from our meta-analytic models as per protocol (Supplementary Results).

From the remaining 48 reports of 45 studies, 26 of 48 (54%) included data on INSTI drug resistance at failure, whereas 22 of 48 (46%) did not, either because no genotype was performed at failure (10/22, 45%) or because no cases of VF were reported (12/22, 55%). Most studies (21/48, 44%) defined VF as 2 consecutive VL measurements of over 200 copies/mL or 1 VL over 200 copies/mL followed by regimen change or discontinuation (Supplementary Table 1). Additional characteristics of study populations are included in Table 1 and Supplementary Table 2.

Table 1.

Outcome Summary Table

Trial	Study Design	Population	Follow-up, wk	Regimen	N	VF (ITT%)	Successful Genotypes	DRM (ITT%) [Geno%]
Induction-maintenance
FLAIR [7]	RCT	Adult PWH	48	4-weekly CAB + RPV	283	4 (1.4)	3	3 (1.1) [100]
LATTE-2 [8]	RCT	Adult PWH	96	Both 4- and 8-weekly CAB + RPV	230	2 (1.0)	2	1 (0.5) [50]
Switch-suppressed
Adachi et al (2024) [20]	Retrospective cohort	Adult PWH	56	8-weekly CAB + RPV	74	0	0	0
Antonucci et al (2024) [67]	Prospective cohort	Adult PWH	55	8-weekly CAB + RPV	79	2 (2.5)	2	0
ATLAS [5]	RCT	Adult PWH	52	4-weekly CAB + RPV	308	3 (1.0)	3	1 (1.0) [33]
ATLAS 2M [6]	RCT	Adult PWH	48	8-weekly CAB + RPV	522	8 (1.5)	7	2 (1.3) [87.5]
				4-weekly CAB + RPV	523	2 (0.4)	2	2 (0.4) [100]
Bailón et al [49]	Case report	Adult PWH	5	4-weekly CAB + RPV	1	1 (100)	1	0
CARES [21]	RCT	Adult PWH	48	8-weekly CAB + RPV	255	2 (0.8)	2	2 (0.8) [100]
Canavasi et al (2024) [66]	Prospective cohort	Adult PWH	23	NA	574	11 (1.9)	10	4 (0.7) [40]
D'Amico et al (2023), compassionate-use program [24]	Prospective cohort	Adult PWH—oral regimen unsuitable	40	4-weekly CAB + RPV	7	1 (14.0)	1	1 (14.0) [100]
Eron et al (2023), TRIO [46]	Retrospective cohort	Adult PWH	40	8-weekly CAB + RPV	278	2 (0.7)	0	0
Fessler et al (2024) [27]	Cross-sectional	Adult PWH	36	8-weekly CAB + RPV	141	19 (13.0)	NA	NA
Gagliardini et al (2024), ICONA [65]	Prospective cohort	Adult PWH	35	NA	470	2 (0.4)	2	1 (0.2) [50]
Gandhi et al (2023), SPLASH [11]	Prospective cohort	Adult PWH—poor adherence	33	4-weekly CAB + RPV	76	0	0	0
Garris et al (2022), CUSTOMIZE [26]	Prospective cohort	Adult PWH	52	4-weekly CAB + RPV	109	0	0	0
Hsu et al (2024), OPERA [32]	Prospective cohort	Adult PWH	24	4-weekly CAB + RPV	1362	25 (2.0)	NA	NA
John et al (2023), JABS [40]	Prospective cohort	Adult PWH—poor adherence	48	8-weekly CAB + RPV	60	0	0	0
Jonsson-Oldenbüttel et al (2024), CARISEL [22]	Prospective cohort	Adult PWH	52	8-weekly CAB + RPV	430	1 (0.2)	1	0
Jonsson-Oldenbüttel et al (2022), CARLOS [73]	Prospective cohort	Adult PWH	24	8-weekly CAB + RPV	207	2 (1.0)	2	1 (0.4) [100]
Konishi et al (2024) [64]	Prospective cohort	Adult PWH	44	8-weekly CAB + RPV	38	0	0	0
Liegeon et al (2024) [38]	Retrospective cohort	Adult PWH	32	8-weekly CAB + RPV	78	1 (1.3)	1	1 (1.3) [100]
Lagi et al (2024), LAHIV [63]	Retrospective cohort	Adult PWH	38	NA	102	1 (1.0)	1	0
Maguire et al (2024) [28]	Retrospective cohort	Adult PWH—high BMI	36	Both 4- and 8-weekly CAB + RPV	374	4 (1.1)	3	2 (0.5) [67.0]
Mesa et al (2023) [41]	Retrospective cohort	Adult PWH	24	CAB + RPV	78	1 (0.1)	NA	NA
Gaur et al (2024), MOCHA [31]	Single-arm	Adolescents and young adults	24	4-weekly CAB + RPV	144	0	0	0
Gersteberg et al (2024) [71]	Case series	Adult PWH	NA	8-weekly CAB + RPV	5	1	1	1 (20) [100]
Montalvo et al (2023), RWHAP [44]	Prospective cohort	Adult PWH	36	Other	30	0 (0)	NA	NA
Muccini et al (2024), SCohorLART [62]	Prospective cohort	Adult PWH	52	8-weekly CAB + RPV	514	4 (0.8)	4	2 (0.4) [50]
Nguyen et al (2024) [42]	Prospective cohort	Adult PWH	17	Both 4- and 8-weekly CAB + RPV	73	3 (4.1)	3	1 (1.4) [100]
Pecora Fulco et al (2024) [50]	Case study	Adult PWH	52	8-weekly CAB + RPV	1	0	0	0
Perez et al (2024), CORE [25]	Retrospective cohort	Adult PWH	24	8-weekly CAB + RPV	62	2 (3.2)	NA	NA
POLAR [39]	nRCT	Adult PWH	52	8-weekly CAB + RPV	90	0	0	0
Pozniak et al (2024) [74], COMBINE-2	Prospective cohort	Adult PWH	13	Both 4- and 8-weekly CAB + RPV	472	3 (0.6)	…	…
Rakhamanina et al (2022) [54]	Case series	Adolescents and young adults	33	Both 4- and 8-weekly CAB + RPV	3	3 (100.0)	2	1 (33.3) [50]
Ring et al (2024), SHARE LAI-net [48]	Retrospective cohort	Adult PWH	32	Both 4- and 8-weekly CAB + RPV	423	3 (0.7)	3	0
Rubenstein (2023) [43]	Prospective cohort	Adult PWH	16	Other	58	1 (2.0)	1	0
Shankaran et al (2024) [45]	Prospective cohort	Adult PWH	104	8-weekly CAB + RPV	75	3 (4.0)	3	3 (4.0) [100]
Senkoro et al (2023) [52]	Case study	Adult PWH—oral regimen unsuitable	4	4-weekly CAB + RPV	1	0	0	0
Serris et al (2024) [72]	Retrospective cohort	Adult PWH	39	8-weekly CAB + RPV	124	5 (4.0)	3	0
Soffritti et al (2024), MODENA [61]	Prospective cohort	Adult PWH	46	NA	74	2 (2.7)	2	2 (2.7) [100]
SOLAR [9]	RCT	Adult PWH	52	8-weekly CAB + RPV	454	3 (0.7)	3	2 (0.4) [67.0]
Taramasso et al (2024), SCOLTA [70]	Prospective cohort	Adult PWH	9	NA	231	2 (0.9)	NA	NA
Valin et al (2024) [29]	Case study	Adult PWH	4	4-weekly CAB + RPV	1	1 (100.0)	1	1 (100) [100]
van der Wekken-Pas et al (2024) [60]	Case study	Pregnant PWH	52	4-weekly CAB + RPV	1	0	0	0
Yokota et al (2024) [53]	Case study	Adult PWH—on R-CHOP therapy for HIV-related lymphoma	4	4-weekly CAB + RPV	1	1	NA	NA
Switch-viremic
Al-Handola et al (2023) [51]	Case study	Adult PWH	16	4-weekly CAB + RPV	1	0	0	0
Brock et al (2024) [57]	Prospective cohort	Adult PWH	52	Both 4- and 8-weekly CAB + RPV	12	0	0	0
Barnett et al (2022) [47]	Case study	Adult PWH	12	4-weekly CAB + RPV	1	0	0	0
Chilton et al (2019) [59]	Case series	Adult PWH	4	4-weekly CAB + RPV	2	0	0	0
Elion et al (2023), TRIO [34]	Retrospective cohort	Adult PWH	16	Both 4- and 8-weekly CAB + RPV	29	1 (3.4)	NA	NA
D'Amico et al (2022), compassionate-use program [24]	Prospective cohort	Adult PWH—oral regimen unsuitable	40	4-weekly CAB + RPV	28	6 (21.0)	6	2 (7.0) [33.0]
Fessler et al (2024) [27]	Cross-sectional	Adult PWH	36	8-weekly CAB + RPV	31	6 (19.0)	NA	NA
Hickey et al (2024), SPLASH [68]	Retrospective cohort	Adult PWH—poor adherence	24	8-weekly CAB + RPV	59	3 (5.1)	3	1 (1.7) [33.0]
Hill et al (2024) [55]	Retrospective cohort	Adult PWH	41	Both 4- and 8-weekly CAB + RPV	144	1	NA	NA
Hsu (2023), OPERA [36]	Prospective cohort	Adult PWH	24	4-weekly CAB + RPV	229	7 (3.0)	NA	NA
Kilcrease et al (2024) [30]	Case series	Adult PWH	10	8-weekly CAB + RPV	3	0	0	0
LATITUDE; ACTG A5359 [12]	RCT	Adult PWH—poor adherence	52	4-weekly CAB + RPV	146	6 (4.1)	5	2 (1.4) [83.0]
Schneider et al (2023), BEYOND [69]	Prospective cohort	Adult PWH	24	Both 4- and 8-weekly CAB + RPV	308	4 (1.3)	4	2 (0.6) [50.0]
Serris et al (2024) [72]	Retrospective cohort	Adult PWH	39	8-weekly CAB + RPV	2	0	0	0
Thoueille et al, SHCS-879 [33]	Prospective cohort	Adult PWH	52	8-weekly CAB + RPV	186	3 (1.6)	NA	NA
Van Bremen et al (2023) [58]	Case series	Adult PWH	NA	8-weekly CAB + RPV	3	0	0	0
van Welzen et al (2024) [35]	Case series	Adult PWH	56	8-weekly CAB + RPV	5	5 (100)	4	4 (80.0) [100]
Yared et al (2023) [37]	Retrospective cohort	Adult PWH	12	Both 4- and 8-weekly CAB + RPV	51	1 (3.0)	1	1 (3.0) [100]

Trials are listed by name unless no trial name is available, in which case the first author’s name was used as the main identifier. VF is presented in absolute numbers and as a percentage proportion in an ITT population. DRM is presented in absolute numbers, as a percentage proportion in an ITT population, and as a percentage proportion at failure in successfully genotyped participants.

Abbreviations: ATLAS, Antiretroviral Therapy as Long Acting Suppression; ATLAS-2M, Antiretroviral Therapy as Long Acting Suppression every 2 Months; BEYOND, Bystander Effects of HIV on the Endothelial Dysfunction and Ageing; CARES, Cabotegravir and Rilpivirine Efficacy and Safety; BMI, body mass index; CAB, cabotegravir; CARLOS, Cohort of Antiretroviral-naïve, long-term, and optimal suppressors; CARISEL, Cabotegravir and Rilpivirine Implementation Study in European Locations; COMBINE-2, Combination Antiretroviral Therapy to Optimize Treatment in HIV-1-infected Individuals with Cardiovascular Disease Study; CORE, Cohort for Research in HI-v Associated Renal Disease; CUSTOMIZE, Cabotegravir plus Rilpivirine long acting in the US to Optimize and Measure Implementation and Experience; DRM, integrase inhibitor–associated drug-resistance mutation; FLAIR, First Long-Acting Injectable Regimen; Geno, successfully genotyped study participants at failure; HIV, human immunodeficiency virus; ICONA, Italian Cohort of Individuals Naïve for Antiretrovirals; INSTI, Integrase Strand Transfer Inhibitor; ITT, intention-to-treat; JABS, Johannesburg Antiretroviral Biological Study; LATITUDE, Long-Acting Therapy to Improve Treatment SUccess in Daily LifE; LATTE, Long-Acting antireTroviral Treatment Enabling; MOCHA, More Options for Children and Adolescents; N, number of participants randomized into treatment arm; NA, genotype not available; nRCT, non–randomized controlled trial; OPERA, Observational Pharmacoepidemiology Research and Analysis; PWH, people with human immunodeficiency virus; R-CHOP, combination chemotherapy consisting of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone; RCT, randomized controlled trial; RPV, rilpivirine; RWHAP, Ryan White HIV/AIDS Program; SCOLTA, Surveillance Cohort Long-term Toxicity Antiretrovirals; SHARE LAI-net, Sexual Health and HIV All East Research Long-acting Injectable network; SOLAR, Switch Onto Long Acting Regimen; SPLASH, Special Program of Long-Acting Antiretrovirals to Stop HIV; TRIO, Trio Health HIV Network cohort; VF, virologic failure.

Two out of 48 (4%) reports were induction-maintenance studies, 35 of 48 (73%) were switch suppressed, and 12 of 48 (25%) were switch viremic (Table 1). Six studies reported separate results for participants suppressed and unsuppressed at trial entry and were therefore included in both switch-viremic and switch-suppressed groups. Supplementary Table 3 outlines the rationale for the classification of each study. Most studies (39/48, 81%) reported results between 24 and 52 weeks of follow-up, with a median follow-up time of 41 weeks.

Risk of Bias

Among randomized studies, 5 of 7 (86%) were low risk of bias and 2 of 7 (14%) raised “some concerns” (Supplementary Figure 1). Among nonrandomized studies, 10 of 38 (26%) were low risk of bias and 23 of 38 (61%) were moderate risk. The main source of bias was incomplete reporting of study methods in available abstract or poster presentations and failure to adjust for confounding factors (Supplementary Figure 2).

Four studies [25, 56, 63, 72] were ranked as high risk of bias and hence excluded from meta-analysis as per protocol: 3 had large proportions of missing outcome data (range: 31%–71%) and the fourth excluded 42% of participants for not having 6 months of VL follow-up, despite all reported cases of VF happening within 4 months of initiating CAB/RPV (Supplementary Figure 2).

Protocol-Defined VF

Thirty-five reports of 33 studies including 9224 participants reported incidence of protocol-defined VF between 24 and 52 weeks. Two of these (2/33, 6%) were induction-maintenance studies, 27 of 33 (82%) switch-suppressed studies, and 9 of 33 (29%) were switch-viremic studies (Table 1).

As shown in Figure 2A, protocol-defined VF was identified in 6 of 513 participants in induction-maintenance studies (meta-analysis pooled estimate: 1%; 95% CI: 1%–3%), 92 of 7801 participants in switch-suppressed studies (1%; 95% CI: 1%–2%), and 37 of 910 participants in switch-viremic studies (5%; 95% CI: 3%–10%).

Figure 2.

A, Single-proportion meta-analysis for prevalence of protocol-defined virologic failure between 24 and 52 weeks for induction-maintenance, switch-suppressed, and switch-viremic studies. B, Single-proportion meta-analysis for prevalence of treatment-emergent INSTI resistance between 24 and 52 weeks for induction-maintenance, switch-suppressed, and switch-viremic studies. Abbreviations: ATLAS, Antiretroviral Therapy as Long Acting Suppression; ATLAS-2M, Antiretroviral Therapy as Long Acting Suppression every 2 Months; BEYOND, Bystander Effects of HIV on the Endothelial Dysfunction and Ageing; CARES, Cabotegravir and Rilpivirine Efficacy and Safety; CARLOS, Cohort of Antiretroviral-naïve, long-term, and optimal suppressors; CARISEL, Cabotegravir and Rilpivirine Implementation Study in European Locations; COMBINE-2, Combination Antiretroviral Therapy to Optimize Treatment in HIV-1-infected Individuals with Cardiovascular Disease Study; CORE, Cohort for Research in HI-v Associated Renal Disease; CUSTOMIZE, Cabotegravir plus Rilpivirine long acting in the US to Optimize and Measure Implementation and Experience; FLAIR, First Long-Acting Injectable Regimen; ICONA, Italian Cohort of Individuals Naïve for Antiretrovirals; INSTI, Integrase Strand Transfer Inhibitor; JABS, Johannesburg Antiretroviral Biological Study; LATITUDE, Long-Acting Therapy to Improve Treatment SUccess in Daily LifE; LATTE, Long-Acting antireTroviral Treatment Enabling; MOCHA, More Options for Children and Adolescents; RWHAP, Ryan White HIV/AIDS Program; OPERA, Observational Pharmacoepidemiology Research and Analysis; SCOLTA, Surveillance Cohort Long-term Toxicity Antiretrovirals; SHARE LAI-net, Sexual Health and HIV All East Research Long-acting Injectable network; SOLAR, Switch Onto Long Acting Regimen; SPLASH, Special Program of Long-Acting Antiretrovirals to Stop HIV; TRIO, Trio Health HIV Network cohort.

Treatment-Emergent INSTI Resistance

Nineteen reports of 19 studies including 5662 total participants and 59 successfully genotyped participants at failure reported resistance data between 24 and 52 weeks. Two of these (2/19, 10%) were induction-maintenance studies, 14 of 19 (75%) switch-suppressed studies, and 4 of 19 (25%) were switch-viremic studies (Table 1).

As shown in Figure 2B, INSTI resistance developed at failure in 4 of 5 participants in induction-maintenance studies (meta-analysis pooled estimate: 71%; 95% CI: 25%–95%), 22 of 37 participants in switch-suppressed studies (61%; 95% CI: 44%–75%), and 7 of 17 participants in switch-viremic studies (41%; 95% CI: 20%–65%).

Individual genotype results were available for 18 of 19 studies. Sixty-four major INSTI-associated DRMs were detected in viral sequences from 33 individuals, 21 of these 33 individuals (64%) also conferred intermediate or high levels of DTG cross-resistance (Supplementary Table 1). The most common INSTI-associated DRMs were as follows: Q148R (17/64, 27%), N155H (11/64, 17%), E138K (8/64, 13%), G140S (8/64, 13%), and Q148K (4/64, 8%).

Sensitivity Analyses

Given that the incidence of events was found to be rare for both VF and resistance analyses, an additional sensitivity analysis applying the Freeman–Tukey transformation was performed. Results did not vary significantly from our main analysis (Supplementary Figure 3).

If we narrowed the follow-up window to 48 ± 4 weeks, point estimates in induction-maintenance, switch-suppressed, and switch-viremic studies remained statistically similar for both rates of VF (1% [95% CI: 1%–3%], 1% [95% CI: 1%–1%], and 4% [95% CI: 2%–7%], respectively) and INSTI resistance (71% [95% CI: 25%–95%], 66% [95% CI: 45%–81%], and 50% [95% CI: 19%–81%], respectively).

Results were consistent in all other sensitivity analyses standardizing or redefining outcome definitions (Supplementary Figure 3, Supplementary Table 4).

Certainty of Evidence

The certainty of evidence was considered moderate for both outcomes (Supplementary Table 5). It was downgraded for indirectness due to high variability in location and follow-up duration, as well as use of strict inclusion criteria, which may not be fully representative of the real-world population of PWH.

Moreover, the certainty of evidence in our resistance analysis was downgraded for imprecision due to small sample sizes, which may have resulted in overreporting of resistance rates. Certainty of evidence for VF was downgraded for publication bias as suggested by an asymmetrical funnel plot (Supplementary Figure 4) and the frequent involvement of study funders in decisions to publish.

DISCUSSION

In a meta-analysis of 33 studies involving 9224 individuals, we found that VF was rare in those taking CAB/RPV for ART, with rates of 1% in induction-maintenance, 1% in switch-suppressed, and 5% in switch-viremic groups. This supports CAB/RPV as a highly effective treatment, especially in settings where adherence monitoring may be challenging, like in low- and middle-income countries (LMICs). However, INSTI resistance among those with failure was common, with rates of 71%, 61%, and 41% in those in induction-maintenance, switch-suppressed, and switch-viremic therapy, respectively (Figure 3). By contrast, a recent meta-analysis by Chu et al [75] assessing the efficacy of oral DTG found resistance rates of 0% in switch-suppressed studies and 23% in switch-viremic studies. Resistance rates in ART-naive individuals were only 1%, even without needing to achieve viral suppression before initiating DTG-based ART, unlike in corresponding CAB/RPV induction-maintenance studies. These findings align with models predicting increased emergence of INSTI resistance with CAB/RPV use [76], underscoring its implications for both individual and broader population resistance trends.

Figure 3.

Bar chart displaying the prevalence of virologic failure (A) and treatment-emergent INSTI resistance (B) in PWH on CAB/RPV ART regimens for induction-maintenance, switch-suppressed, and switch-viremic studies. Rates of virologic failure are calculated in an intention-to-treat population, whereas resistance emergence is calculated as a proportion of successfully genotyped participants at failure. Abbreviations: ART, antiretroviral therapy; CAB, cabotegravir; INSTI, integrase strand transfer inhibitor; PWH, people with human immunodeficiency virus; RPV, rilpivirine.

Genetic mutations in those experiencing CAB/RPV failure often led to high levels of DTG cross-resistance. While the high genetic barrier to resistance of DTG has been widely documented for those newly initiating therapy or without prior INSTI exposure [76], its efficacy drops significantly in those previously exposed to INSTIs [1, 77]. With 91% of countries using DTG-based therapies as first-line ART in 2024, and 77% including DTG in second-line regimens [1], the potential spread of INSTI-resistant HIV strains through wider CAB/RPV implementation, particularly in high-risk individuals with adherence challenges, could impact the durability of INSTI-based regimens [78]. Reassuringly, VF rates seen for CAB/RPV therapy compared favorably with those calculated for DTG-based therapies by Chu et al [75], where VF rates were 3% and 11% for switch-suppressed and switch-viremic, respectively, and 5% in ART-naive individuals. Long-Acting Therapy to Improve Treatment SUccess in Daily LifE (LATITUDE) [12] and Cabotegravir and Rilpivirine Efficacy and Safety (CARES) [21], the only 2 trials comparing DTG and CAB head-to-head, similarly found higher viral suppression rates in the CAB/RPV groups.

These differences could be attributable to the long-acting formulation and hence reduced dosing schedule of injectable CAB/RPV compared with daily oral standard of care, helping overcome stigma and adherence challenges related to pill burden [4]. It is uncertain, however, if this benefit offsets the higher risk of treatment-emergent resistance [15], especially in real-world settings where monitoring of adherence and VL is harder. Injectable therapies hold the potential to benefit vulnerable populations, but more data are needed outside of controlled trials to confirm these effects.

Notwithstanding the high virologic suppression rates with CAB/RPV, without new licensing agreements, the current price of patented long-acting injectable CAB/RPV in high-income settings remains a major barrier to treatment in countries with the highest burdens of HIV, where oral DTG-based ART costs under US $45 per person annually, compared to over US $23 000 for injectables [79]. Moreover, since drug prices and healthcare costs often influence willingness for HIV treatment among trial participants [2], additional expenditures associated with long-acting ART could deter policymakers and potential users alike.

Limitations

As with all meta-analytic exercises, our results should be interpreted within the context of included studies. For example, most studies excluded participants with detectable viremia and/or prior INSTI resistance, which could mean our review overestimates the real-world efficacy of CAB/RPV, especially in low-resource settings where VL monitoring and genetic sequencing are not always feasible. Because VF is rare with CAB/RPV, we encountered relatively few reports of emergent resistance, especially in those initiating CAB/RPV with viremia. Although switch-viremic studies included 4%–17% viremic participants, studies did not always stratify results by pretreatment VL. Only 6 studies [12, 24, 27, 32, 34, 80] and 3 case reports [29, 30, 47] reported VF rates in fully viremic populations. Resistance data at failure were only available for 1 case series [35] and 3 observational studies [12, 37, 80]. Ultimately, additional data are needed to clarify resistance risks in viremic individuals switching to CAB/RPV, who may benefit most from this regimen.

Our analysis overrepresented PWH with assigned male sex at birth (75% of participants) and populations in high-income settings (only CARES [21] and More Options for Children and Adolescents [MOCHA] [31] studies were set in LMICs). Since over 80% of new HIV infections occur in LMICs and closer to 50% of PWH globally are women and girls [81], more data on CAB/RPV efficacy in these populations are essential. Many studies excluded PWH with the HIV-1 A6 subtype—a strong predictor of VF [80, 81]. Although still rare globally, the prevalence of HIV-1 A6 is increasing because of migration [82], potentially impacting the future efficacy of CAB/RPV and hence real-world applicability of our results.

Only half of the studies included performed resistance testing at failure and, among those that did, 7% of genotyping sequences were not successful. Because our analysis defined resistance only on successful genotypes, it effectively excludes a small proportion of potentially informative results, possibly affecting resistance estimates. However, when using total number of VF cases as the denominator, results remained similar, increasing our confidence that missing data were not dependent on real results.

Another limitation of this meta-analysis is the small sample sizes in most studies identified, particularly in our resistance analysis, leading to point estimates with wide CIs. Despite using inverse-variance weighted percentages to account for these differences in study variability, there is still a risk that findings may not extrapolate to larger populations.

Follow-up periods for individual studies were also relatively short (median: 41 weeks) and heterogenous (range: 24 to 52 weeks). This may have resulted in our analysis underreporting failure and resistance rates or conclusions being too premature. However, sensitivity analyses using varying outcome window periods showed consistent results. Moreover, data from RCTs show minimal changes in VF rates and resistance emergence over a span of 48 to 152 weeks [83]. Although this increases confidence that our conclusions would hold over time, larger real-world cohort studies with longer follow-ups are needed for further validation.

Conclusion

In conclusion, we found low rates of VF among PWH taking long-acting injectable CAB/RPV, but rates of treatment-emergent INSTI resistance among those experiencing failure were high, particularly when compared with cohorts with VF on oral DTG-based regimens. Thus, individual- and population-level surveillance of emerging resistance in populations taking CAB/RPV may be warranted.

Supplementary Data

Supplementary materials are available at Clinical 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.

Notes

Author contributions. A. P. N. led the study conception, literature search and screening, author correspondence, data collection, results analysis and interpretation, and manuscript preparation. M. J. S., S. M. M., and A. H. contributed to the study conception and critical review of the manuscript. C. T. N. assisted with title and abstract screening and critical manuscript review.

Financial support. This work was supported by the World Health Organization (grant number 2024_MV_001 to A. H.) and the National Institutes of Health (grant numbers T32 AI007387 [to C. T. N.], K24 HL166024 [to M. J. S.], and K23 AI143470 [to S. M. M.]). The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Data availability. The data that support the findings of this study are available from the corresponding author, A.P.N., upon reasonable request.