FormalPara Key Summary Points

Why carry out this study?

Long-acting bronchodilators are the mainstay of maintenance therapy for patients with chronic obstructive pulmonary disease (COPD), but there is no consensus on the timing of treatment initiation with dual therapy with a long-acting muscarinic antagonist and a long-acting β2-agonist (LAMA/LABA) combination versus monotherapy with a LAMA or a LABA.

Several dual bronchodilator therapies are available for COPD treatment; however, only a few head-to-head randomised controlled trials have compared outcomes between dual bronchodilator therapies, and these may have been affected by differences in inhaler devices.

This network meta-analysis investigated the relative efficacy of umeclidinium/vilanterol (UMEC/VI) versus other dual bronchodilator combinations and LAMA and LABA monotherapies with reduced confounding due to inhaler device differences.

What was learned from the study?

UMEC/VI dual therapy provided better outcomes in terms of lung function (as measured by change from baseline in trough FEV1) compared with alternative dual therapies and monotherapies, as well as improvements in health-related quality of life, symptoms, rescue medication use, moderate/severe exacerbation rates, and time to first moderate/severe exacerbation compared with monotherapies.

These results suggest that treatment with UMEC/VI may improve outcomes for symptomatic patients with COPD compared with alternative dual and monotherapies.

Introduction

Long-acting bronchodilators are the mainstay of maintenance therapy for patients with chronic obstructive pulmonary disease (COPD). However, treatment guidelines provide different recommendations on when to initiate long-acting muscarinic antagonist (LAMA)/long-acting β2-agonist (LABA) dual therapy rather than LAMA or LABA monotherapy. The Global Initiative for Chronic Obstructive Lung Disease (GOLD) strategy report typically recommends a stepwise approach with escalation to LAMA/LABA dual therapy for patients who have persistent dyspnoea or exacerbations on LAMA or LABA monotherapy, although LAMA/LABA dual therapy can also be considered as an initial maintenance therapy option for patients with severe symptoms [1]. In contrast, the American Thoracic Society (ATS) and the UK National Institute for Health and Care Excellence (NICE) guidelines recommend initiating treatment with LAMA/LABA dual therapy for all patients with dyspnoea [2, 3].

Previous meta-analyses of clinical trial data comparing dual and mono-bronchodilator therapies have shown that treatment with LAMA/LABA combinations provide better outcomes for patients than LAMA or LABA monotherapy [4,5,6,7,8,9]. All currently available LAMA/LABA combinations provided greater improvements in lung function, health status and breathlessness compared with monotherapies [4,5,6, 9]. However, further evidence from large randomised controlled trials (RCTs) evaluating efficacy and safety of LAMA/LABA dual therapy versus monotherapy is now available, so there is a need for updated analyses to incorporate these findings.

The relative differences in safety and efficacy between available LAMA/LABA treatments within the dual bronchodilator class also remain unclear. There is a need to identify whether there is a difference between bronchodilators in the benefits they produce across different outcomes to help guide treatment decisions. Previous meta-analyses showed a gradient of effectiveness between LAMA/LABA fixed-dose combinations in patients with moderate-to-severe COPD based on the evidence available in 2015 [4, 6]. Although within-class RCTs comparing outcomes between LAMA/LABA dual therapies remain limited, recent evidence may help to clarify the relative benefits of different LAMA/LABA combinations.

An updated synthesis of recent and previously available evidence is therefore needed to understand the potential advantages of starting treatment with LAMA/LABA dual therapy versus LABA or LAMA monotherapy, and which treatments provide the best outcomes within the LAMA/LABA dual therapy class. The aim of this network meta-analysis (NMA) was to investigate whether treatment with umeclidinium/vilanterol (UMEC/VI) provides better outcomes than (a) monotherapy with LABA or LAMA, and (b) other dual bronchodilator combinations in symptomatic patients with moderate-to-severe COPD.

Methods

Systematic Literature Review Rationale

The systematic literature review (SLR) methodology was consistent with recommendations published in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement, as well as guidance provided by the Centre for Reviews and Dissemination and the Cochrane Collaboration [10, 11].

A previously published SLR identified publications up to October 2015 reporting RCTs in patients with COPD that compared selected maintenance therapies, including LAMA and LABA monotherapies and LAMA/LABA dual therapies, to inform an NMA [6]. An update to this search was conducted to identify publications between October 2015 and November 2020 reporting studies in which any LAMA, LABA or LABA/LAMA either as a monotherapy or as a combination with inhaled corticosteroids (ICS; dual therapy) were compared with LABA/LAMA dual therapy (fixed-dose or open combination). Studies with an ICS/LABA treatment arm were used to construct the networks, but were excluded from the comparisons.

Search Strategy

The systematic bibliographic searches were conducted between 2 October 2015 and 19 November 2020 using the Ovid Platform in the following databases: MEDLINE and MEDLINE In-Process, EMBASE, The Cochrane Library: Cochrane Database of Systematic Review (CDSR) and Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE), Health Technology Assessment (HTA) websites, HTA database and the National Institute for Health Research (NIHR). To complement the evidence from the bibliographic databases, a secondary systematic searches were performed in clinical trial registries including Clinicaltrials.gov (https://clinicaltrials.gov/ct2/search/advanced), the US National Institutes of Health clinical trial register; World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; http://apps.wo.int/trialsearch/AdvSearch.aspx); ISRCTN registry Clinical Trials Register (ISRCTN; http://www.controlled-trials.com/editAdvancedSearch); Klinische Prüfungen PharmNet.Bund (http://www.pharmnet-bund.de/dynamic/de/klinische-pruefungen/index.htm); the International Prospective Register of Systematic Reviews (PROSPERO; https://www.crd.york.ac.uk/prospero/#searchadvanced); National Institute for Health Research—Health Technology Assessment (NIHR HTA; http://www.nets.nihr.ac.uk/projects) and EU Clinical Trials Register (EU-CTR; www.clinicaltrialsregister.eu). The search was restricted to articles written in English or German.

Inclusion Criteria and Study Selection Process

The outcomes included in the SLR are listed in Table 1. For all outcomes evaluated at 12 and 24 weeks, an extended margin of time was permitted to strengthen the networks. If the endpoint of interest was not reported at 12 or 24 weeks, but instead between 8 and 16 weeks and 20 and 28 weeks, the proxy outcome was reported. The primary outcome was trough FEV1 at 24 weeks.

Table 1 PICOS criteria

All abstracts and full articles were reviewed against the eligibility criteria by two systematic reviewers; disagreements were referred to a third reviewer and an agreement was reached. A PRISMA flow diagram, indicating the numbers of studies included and excluded at each stage of the review, is shown in Fig. 1.

Fig. 1
figure 1

PRISMA flow diagram of study selection process. CSR clinical study report, NMA network meta-analysis, SLR systematic literature review

Quality Assessment

The Cochrane Collaboration’s tool was used to assess risk of bias at study level. The scored items were extracted in the data extraction form.

Network Meta-Analysis

In traditional pairwise meta-analysis, all included studies compare the same intervention with the same comparator. NMA is an extension of this approach that includes multiple different pairwise comparisons across a range of different interventions (Supplementary Methods) [12,13,14]. For the current study, all analyses were conducted using a frequentist weighted regression-based approach as previously described [15]. The efficacy outcomes were lung function (trough forced expiratory volume in 1 s [FEV1]), health-related quality of life (HRQoL; St George’s Respiratory Questionnaire [SGRQ] total score), breathlessness (Transitional Dyspnoea Index [TDI] focal score), rescue medication use, moderate/severe exacerbation rate and time to first moderate/severe exacerbation. For safety outcomes, analyses were not feasible because of differences in outcome definitions. Both fixed effects (FE) and random effects (RE) models were used; in the absence of heterogeneity, results from both models were identical. Where heterogeneity was present, the RE models automatically accounted for this; RE model results are presented in the supplementary material. Networks were stratified by observation time horizon (12 and 24 weeks) depending on data availability and the primary comparison was at 24 weeks. Results from the frequentist approach are presented as point estimates with a 95% confidence interval (CI). Point estimates and 95% CIs are presented in figures and tables; only statistically significant results at 24 weeks have been reported in the text. The frequentist regression-based NMA was conducted using R v4.1.2 (www.r-project.org), using the netmeta package [16]. Further details are provided in the Supplementary Methods.

Compliance with Ethics Guidelines

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Results

Studies and Patient Characteristics

In the SLR and its update, a total of 6847 abstracts were reviewed following the removal of duplicates. Of those, 753 full length publications were assessed for inclusion and data were extracted from 83 publications and 13 GSK clinical study reports (CSRs). In total, 96 publications reporting 61 studies were included in the SLR. Following assessment of the feasibility of including studies identified in the SLR, the NMA included 69 publications (including 10 GSK CSRs) reporting 49 different studies (Fig. 1).

The characteristics of the studies included in the NMA are shown in Table 2; the characteristics of patients included in these studies are shown in Table 3.

Table 2 Study characteristics of the trials included in the NMA
Table 3 Baselines characteristics of patients in the studies included in the NMA

Lung Function

Trough FEV1 data were available from 22 and 44 studies at 24 and 12 weeks, respectively (Supplementary Fig. S1). Not all comparators had data available at 24 weeks, including tiotropium/olodaterol (TIO/OLO) fixed-dose and open combinations. The networks of evidence showed that the TONADO 1 and TONADO 2 studies formed a disconnected network at 24 weeks (Supplementary Fig. S1A). In the FE model at 24 weeks, change from baseline in trough FEV1 was statistically significant in favour of UMEC/VI versus all comparators, including dual therapies (Fig. 2a) and monotherapies (Fig. 2b). Mean difference in change from baseline in trough FEV1 (95% CI) versus dual therapies was 108.87 ml (79.27, 138.47) versus aclidinium/formoterol (ACL/FOR) 400/6 µg, 97.50 ml (72.89, 122.11) versus ACL/FOR 400/12 µg, 74.19 ml (55.06, 93.31) versus glycopyrronium (GLY)/FOR 18/9.6 µg, 39.43 ml (19.56, 59.30) versus indacaterol (IND)/GLY 110/50 µg, and 107.43 ml (70.18, 144.67) versus TIO 18 µg + FOR 12 µg. UMEC/VI showed clinically meaningful improvements (treatment difference ≥ 100 ml) [17] compared with ACL/FOR 400/6 µg and TIO 18 µg + FOR 12 µg dual therapies. At 12 weeks, UMEC/VI provided statistically significantly greater improvements in trough FEV1 than all dual therapies, with the exception of GLY/IND (Supplementary Fig. S2A), and provided greater improvements than all monotherapies and placebo (Supplementary Fig. S2B). All therapies were significantly more effective than placebo in increasing trough FEV1, with UMEC/VI providing the largest improvements at both time points (Supplementary Figs. S3 and S4). The RE model produced consistent results (Supplementary Tables S1 and S2).

Fig. 2
figure 2

Fixed effects model of mean difference in change from baseline in trough FEV1 of UMEC/VI versus a dual therapy and b monotherapy at 24 weeks. Assessment of heterogeneity/inconsistency: I2 = 35.33%; Q = 44.84, p = 0.0305. ACL aclidinium, CFB change from baseline, CI confidence interval, FEV1 forced expiratory volume in 1 s, FOR formoterol fumarate, FP fluticasone propionate, GLY glycopyrronium, IND indacaterol, PBO placebo, SAL salmeterol, TIO tiotropium, UMEC umeclidinium, VI vilanterol

Health-Related Quality of Life

SGRQ total score was available in 17 studies at 24 weeks (Supplementary Fig. S5) and 20 studies at 12 weeks. In the FE model at 24 weeks, there was no evidence of any statistically significant differences in SGRQ total score with UMEC/VI compared with all other LAMA/LABAs (Fig. 3a), whereas UMEC/VI provided statistically significantly greater improvements versus UMEC 125 μg (mean difference in change from baseline [95% CI] − 1.89 [− 3.49, − 0.30), GLY 18 μg (− 2.12 [− 3.61, − 0.62]), GLY 50 μg (− 1.43 [− 2.67, − 0.18]), TIO 18 μg (− 1.37 [− 2.42, − 0.32]), and salmeterol (SAL) 50 μg (− 1.79 [− 3.04, − 0.54]), but not versus other monotherapies (Fig. 3b). At 12 weeks, improvements in SGRQ total score with UMEC/VI were not statistically significant compared with other LAMA/LABA combinations; statistically significant improvements were seen for IND/GLY 110/50 μg and TIO 18 µg + IND 150 µg compared with UMEC/VI (Supplementary Table S2). UMEC/VI provided statistically significantly greater improvements in SGRQ total score at 12 weeks than TIO 18 μg and SAL 50 μg, but not compared with the other monotherapies (Supplementary Table S2). In the FE model, all treatments provided statistically significant improvements in SGRQ total score versus placebo, with the exception of UMEC 125 μg at 24 and 12 weeks and SAL 50 μg and IND 150 μg at 24 weeks (Supplementary Tables S3 and S4).

Fig. 3
figure 3

Fixed effects model of mean difference in change from baseline in SGRQ total score of UMEC/VI versus a dual therapy and b monotherapy at 24 weeks. Assessment of heterogeneity/inconsistency: I2 = 22.49%; Q = 32.25; p = 0.1508. ACL aclidinium, CFB change from baseline, CI confidence interval, FOR formoterol fumarate, FP fluticasone propionate, GLY glycopyrronium, IND indacaterol, PBO placebo, SAL salmeterol, SGRQ St George’s Respiratory Questionnaire, TIO tiotropium, UMEC umeclidinium, VI vilanterol

SGRQ responder analyses were available in 14 and 12 studies at 24 (Supplementary Fig. S6) and 12 weeks, respectively. At 24 weeks, UMEC/VI was associated with a statistically significantly greater proportion of SGRQ responders compared with UMEC 62.5 μg (SGRQ responders odds ratio [95% CI] 1.19 [1.02, 1.40]), GLY 18 μg (1.54 [1.22, 1.95]), TIO 18 μg (1.18 [1.00, 1.39]), FOR 9.6 μg (1.36 [1.08, 1.72]) and SAL 50 μg (1.47 [1.22, 1.78]) (Supplementary Fig. S7). At 12 weeks, the odds of being a responder were significantly greater with UMEC/VI versus UMEC 62.5 μg, ACL 400 μg and SAL 50 μg (Supplementary Table S2). All treatments provided statistically significantly greater proportion of SGRQ responders compared with placebo, with the exception of GLY 18 μg at 24 weeks; ACL 400 μg and FOR 12 μg at 12 weeks; and SAL 50 μg at both time points (Supplementary Tables S3 and S4).

The RE model produced consistent results (Supplementary Tables S1–S4).

Breathlessness

Breathlessness as measured by TDI focal score was available in 14 studies at 24 weeks (Supplementary Fig. S8) and 21 studies at 12 weeks. In the FE model at 24 weeks, TDI focal score was statistically significant in favour of UMEC/VI versus GLY/FOR 18/9.6 µg (mean difference in change from baseline [95% CI] 0.33 [0.13, 0.52]), UMEC 62.5 µg (0.32 [0.08, 0.57]), UMEC 125 µg (0.55 [0.16, 0.93]), GLY 18 µg (0.68 [0.32, 1.04]), TIO 18 µg (0.34 [0.03, 0.64)], VI 25 µg (0.42 [0.13, 0.71]), FOR 9.6 µg (0.48 [0.11, 0.84]) and SAL 50 µg (0.43 [0.14, 0.72]) (Fig. 4). At 12 weeks, UMEC/VI provided statistically significantly greater improvements in TDI focal score than UMEC 62.5 µg, UMEC 125 µg, TIO 18 µg, VI 25 µg and SAL 50 µg; IND/GLY 27.5/15.6 µg, TIO/OLO 2.5/5 µg and TIO/OLO 5/5 µg provided statistically significantly greater improvements in TDI focal score than UMEC/VI (Supplementary Table S2). At both time points, all therapies provided statistically significantly greater improvements than placebo, with the exception of TIO 18 µg + FOR 12 µg at 12 weeks (Supplementary Tables S3 and S4).

Fig. 4
figure 4

Fixed effects model of mean difference in change from baseline in TDI focal score of UMEC/VI versus a dual therapy and b monotherapy at 24 weeks. Assessment of heterogeneity/inconsistency: I2 = 0%; Q = 12.60; p = 0.4793. ACL aclidinium, CFB change from baseline, CI confidence interval, FOR formoterol fumarate, FP fluticasone propionate, GLY glycopyrronium, IND indacaterol, PBO placebo, SAL salmeterol, TDI Transitional Dyspnoea Index, TIO tiotropium, UMEC umeclidinium, VI vilanterol

TDI responder analyses were available for 10 and 11 studies at 24 (Supplementary Fig. S9) and 12 weeks, respectively. In the FE model, the odds of being a responder were statistically significantly greater with UMEC/VI versus GLY 15.6 µg and TIO 18 µg at 12 weeks, and versus UMEC 62.5 µg, VI 25 µg and SAL 50 µg at both 24 and 12 weeks (TDI responders odds ratio at 24 weeks [95% CI]; UMEC 62.5 µg: 1.33 [1.14, 1.56]; VI 25 µg: 1.38 [1.14, 1.67]; SAL 50 µg: 1.41 [1.17, 1.70]) (Supplementary Fig. S10; Supplementary Table S2). At both time points, all therapies provided statistically significantly greater proportion of TDI responders than placebo, with the exception of TIO 18 µg + FOR 12 µg at 24 weeks (Supplementary Tables S3 and S4).

The RE model produced consistent results (Supplementary Tables S1–4).

Rescue Medication Use

Rescue medication use data were available in 14 studies at 24 weeks and 15 studies at 12 weeks. At 24 weeks, the ILLUMINATE study was disconnected from the network (Supplementary Fig. S11). In the FE model at 24 weeks, change from baseline in rescue medication use was statistically significant in favour of UMEC/VI versus ACL/FOR 400/12 µg (mean difference in change from baseline [95% CI] − 0.46 [− 0.66, − 0.25]) and all monotherapies (mean difference in change from baseline [95% CI]; UMEC 62.5 µg: − 0.33 [− 0.48, − 0.18]; UMEC 125 µg: − 0.36 [− 0.72, − 0.01]; ACL 400 µg: − 0.37 [− 0.61, − 0.12]; GLY 18 µg: − 0.58 [− 0.80, − 0.37]; GLY 50 µg: − 0.86 [− 1.24, − 0.48]; TIO 18 µg: − 0.50 [− 0.51, − 0.49]; FOR 9.6 µg: − 0.27 [− 0.49, − 0.06]; FOR 12 µg: − 0.27 [− 0.53, 0.00]; SAL 50 µg: − 0.28 [− 0.43, − 0.13]; IND 150 µg: − 0.51 [− 0.89, − 0.13]), with the exception of VI 25 µg: − 0.29 (− 0.64, 0.06) (Fig. 5). At 12 weeks, UMEC/VI provided statistically significantly greater improvements in rescue medication use than TIO/OLO 5/5 µg, UMEC 62.5 µg, TIO 18 µg and SAL 50 µg; IND/GLY 27.5/15.6 µg and TIO 18 µg + IND 150 µg provided statistically significantly greater improvements in rescue medication use than UMEC/VI (Supplementary Table S2). All treatments provided statistically significantly greater improvements in rescue medication use compared with placebo, with the exception of GLY 50 µg at 24 weeks and UMEC 125 µg at 12 weeks (Supplementary Tables S3 and S4).

Fig. 5
figure 5

Fixed effects model of mean difference in change from baseline in rescue medication use of UMEC/VI versus a dual therapy and b monotherapy at 24 weeks. Assessment of heterogeneity/inconsistency: I2 = 49.21%; Q = 33.47; p = 0.0098. ACL aclidinium, CFB change from baseline, CI confidence interval, FOR formoterol fumarate, FP fluticasone propionate, GLY glycopyrronium, IND indacaterol, PBO placebo, SAL salmeterol, TIO tiotropium, UMEC umeclidinium, VI vilanterol

Some differences were observed in the findings of the RE model for rescue medication use. At 24 weeks, change from baseline in rescue medication use was statistically significant in favour of UMEC/VI versus ACL/FOR 400/12 µg, ACL 400 µg, GLY 18 µg, GLY 50 µg and TIO 18 µg (Supplementary Table S1). At 12 weeks, UMEC/VI provided statistically significantly greater improvements versus TIO 18 µg only (Supplementary Table S2).

Annualised Moderate/Severe Exacerbations Rates

Moderate/severe exacerbation data were available in nine studies (Supplementary Fig. S12). In the FE model, UMEC/VI provided statistically significantly lower annualised rates of moderate/severe exacerbations versus ACL/FOR 400/6 µg (incidence rate ratio [95% CI] 0.43 [0.20, 0.91]) and ACL/FOR 400/12 µg (0.43 [0.22, 0.84]), and versus all monotherapies with the exception of UMEC 62.5 µg and TIO 18 µg (Fig. 6). Only UMEC/VI 62.5/25 µg, GLY/FOR 18/9.6 µg and IND/GLY 110/50 µg provided statistically significantly lower annualised rates of moderate/severe exacerbations versus placebo, with IND/GLY 110/50 µg providing statistically significant improvements only in the RE model (Supplementary Table S3).

Fig. 6
figure 6

Fixed effects model of incidence ratio of annualised moderate/severe exacerbations rates of UMEC/VI versus a dual therapy and b monotherapy at 24 weeks. Assessment of heterogeneity/inconsistency: I2 = 38.86%; Q = 6.54; p = 0.1622. ACL aclidinium, CFB change from baseline, CI confidence interval, FOR formoterol fumarate, FP fluticasone propionate, GLY glycopyrronium, IND indacaterol, PBO placebo, SAL salmeterol, TIO tiotropium, UMEC umeclidinium, VI vilanterol

Time to First Exacerbation

In total, 12 studies had data available for time to first exacerbation (Supplementary Fig. S13). In the FE model, hazard of a first exacerbation was statistically significantly lower with UMEC/VI versus SAL 50 µg (hazard ratio [95% CI] 0.64 [0.50, 0.81]) and placebo (0.47 [0.32, 0.67]) (Fig. 7), which was consistent with the RE model (Supplementary Table S1). In the FE model, among dual therapies, UMEC/VI 62.5/25 µg, IND/GLY 110/50 µg and GLY/FOR 18/9.6 µg provided statistically significantly lower hazard of a first exacerbation than placebo, with IND/GLY 110/50 µg only providing statistically significant improvements versus placebo in the RE model (Supplementary Table S3). UMEC 62.5 µg, UMEC 125 µg and TIO 18 µg monotherapies provided statistically significant reductions in hazard of a first exacerbation versus placebo; results from the RE model were similar (Supplementary Table S3).

Fig. 7
figure 7

Fixed effects model of hazard ratio of time to first exacerbation of UMEC/VI versus a dual therapy and b monotherapy at 24 weeks. Assessment of heterogeneity/inconsistency: I2 = 35.54%; Q = 10.86; p = 0.1448. CI confidence interval, FOR formoterol fumarate, FP fluticasone propionate, GLY glycopyrronium, HR hazard ratio, IND indacaterol, PBO placebo, SAL salmeterol, TIO tiotropium, UMEC umeclidinium, VI vilanterol

Safety Overview

Safety outcomes were not analysed in the NMA, but are summarised for the included studies in Supplementary Table S5. Briefly, the percentage of patients with ≥1 adverse event (AE) ranged from 21% to 94%, and the percentage with ≥1 serious AE from 0.6% to 24.0%. Pneumonia was reported in 0–16.0% of patients. In total, 0.2–33.0% of patients withdrew, and 0.5–10.6% of early withdrawals were due to AEs. On-treatment mortality was reported for 0–3.1% of patients.

Discussion

This study evaluated the comparative efficacy of UMEC/VI dual therapy versus other mono and dual therapies in symptomatic patients with COPD at 24 and 12 weeks. Treatment with dual therapy resulted in greater improvements in lung function as measured by trough FEV1 compared with LABA or LAMA monotherapy, with improvements in HRQoL, symptoms, rescue medication use and moderate/severe exacerbation rates seen in some comparisons; AEs were also summarised and showed a similar safety profile across the treatments of interest.

At both time points, all treatments provided greater improvements in lung function compared with placebo; the two UMEC formulations performed better than other monotherapies, and the twice-daily LABAs (SAL 50 µg; FOR 12 µg and FOR 9.6 µg) performed worse than other LABAs. Treatment with UMEC/VI resulted in significant improvements in lung function as measured by trough FEV1 compared with all LAMA and LABA monotherapies. These findings support existing evidence from clinical trials that dual therapy provides greater improvements than monotherapies [18,19,20,21,22], and suggest that starting treatment with dual therapy may improve outcomes for many symptomatic patients compared with a stepwise approach to treatment escalation [1]. The non-interventional prospective DETECT study in 3653 patients showed improvements in lung function, QoL and morning symptoms in the year following a switch from prior treatment (most commonly bronchodilator monotherapy) to fixed-dose ACL/FOR, GLY/IND or UMEC/VI [23]. Previous NMAs have also demonstrated that LAMA/LABA dual therapy is more effective than LAMA or LABA monotherapy in terms of lung function improvement [4, 5, 8]. Notably, while many trials comparing dual bronchodilator therapies with monotherapy are likely to have been affected by the confounding effects of concomitant ICS use, the EMAX trial has demonstrated that UMEC/VI provides better outcomes than UMEC and SAL in symptomatic patients at low exacerbations risk not taking concomitant ICS [22]. Taken together, this evidence suggests that dual therapy may be an appropriate initial maintenance therapy in this patient population, as recommended by the ATS and NICE guidelines [2, 3]. However, it should be noted that the present study also supported the efficacy of LAMA and LABA monotherapies compared with placebo.

At both time points, UMEC/VI demonstrated larger improvements in lung function than other dual therapies. The smallest treatment differences were seen with UMEC/VI versus IND/GLY formulations, while the other dual therapies performed similarly; clinically meaningful differences (based on a mean difference in change from baseline of ≥ 100 ml) [17] were seen with UMEC/VI versus some dual therapies at 24 weeks. Treatment effects on SGRQ total score, TDI focal score, rescue medication use and moderate/severe exacerbations were less clear. The results of this updated NMA are consistent with head-to-head trials. The AERISTO trial in 1119 symptomatic patients with moderate-to-very severe COPD found that GLY/FOR was non-inferior to UMEC/VI when comparing treatment effects on peak FEV1, but was inferior when comparing the co-primary endpoint of trough FEV1 [24]. UMEC/VI provided numerical improvements compared with GLY/FOR on symptoms outcomes (TDI focal score and COPD Assessment Test score) in the AERISTO trial, although these were not considered clinically significant [24]. Another head-to-head study showed that UMEC/VI provided statistically significant increases in lung function-related outcomes compared with TIO/OLO in 236 symptomatic patients [25]. Results for patient-reported outcomes (PRO) were less clear; while UMEC/VI provided a significantly greater reduction in rescue medication use during the study and a significant decrease in CAT score at week 4 versus TIO/OLO, there were no consistent treatment differences on exacerbations or other symptoms outcomes, although the study was not powered to demonstrate treatment effects on these measures [25].

Despite the efficacy gradient suggested by the findings of head-to-head trials, previous NMAs have shown mixed results when comparing treatments within the LAMA/LABA class. An NMA of studies in moderate-to-very severe patients with COPD found that TIO/OLO 5/5 µg was ranked highest on efficacy and cardiovascular safety when compared with alternative LAMA/LABAs, including UMEC/VI 62.5/25 µg; however, SGRQ and TDI response rates, rescue medication use, and moderate/severe exacerbations were not included in the analysis [26]. Another previous NMA including RCTs of patients with stable COPD compared six LAMA/LABA dual therapies, and reported that UMEC/VI reduced total exacerbation events compared with alternative LAMA/LABAs (with the exception of GLY/FOR) in one network, although no statistically significant differences were seen in an alternative network [19]. Similarly, the current updated NMA found that UMEC/VI reduced the rate of moderate/severe exacerbations versus two formulations of ACL/FOR, but not GLY/FOR 18/9.6 µg or IND/GLY 110/50 µg. Taken together, the evidence from head-to-head trials and NMAs generally supports a gradient of effectiveness within the LAMA/LABA class, with UMEC/VI providing greater improvements in lung function compared with other LAMA/LABAs, and improvements in other outcomes including exacerbations and PROs in some analyses. The relationship between improvement in lung function and PROs remains unclear; a pooled analysis of 23 clinical trials in COPD found a correlation between improvement in lung function and PRO improvements [27], but disparities have been noted in other studies [28, 29]. This dissociation may be related to factors such as study design, patient population, and statistical power.

This NMA is an updated extension of our original NMA comparing the efficacy of UMEC/VI with fixed-dose and open LAMA/LABA combinations, with TIO, and with placebo [6]. In the present study we have used well-established frequentist NMA approach to compare UMEC/VI with dual therapies and monotherapies, and we have incorporated a larger number of studies, with an additional five RCTs included in the updated NMA. The updated NMA also included additional outcomes such as the rate of moderate/severe exacerbations, time to first moderate/severe exacerbation, and responder analyses for symptoms and health status outcomes. Similar to our original NMA [6] and to a previous NMA that compared RCTs of patients with COPD comparing five combinations of LAMA/LABA dual therapies [4], the present study also found evidence for a potential gradient of effectiveness in treatment effects within the LAMA/LABA class. However, this updated NMA did not analyse certain outcomes relevant to patients with COPD, such as exercise capacity, because of lack of sufficient clinical trial data to construct networks.

A general strength of NMA approaches is the inclusion of a broader population than individual RCTs; specific strengths of this study include the greater number of studies available for inclusion in the NMA compared with similar previous analyses, which allowed for a better estimation of relative treatment effects. Potential general limitations of NMA approaches include a risk of unknown imbalances in effect modifiers and residual confounding bias. Only RCTs were included in this study, eliminating the risk of bias due to imbalanced effect modifiers within the trials as a result of randomisation. However, randomisation does not control for known treatment effect modifiers across studies, and unknown treatment effect modifiers could still bias the results. Another potential general limitation of NMA approaches is the risk of violation of the three assumptions of similarity, consistency and transitivity. In this study, the similarity assumption (i.e. the comparability of baseline characteristics) and consistency assumption (tested through heterogeneity statistics such as I2, the Q statistic and the corresponding p value) were deemed to hold, providing certainty that the transitivity assumption also holds. This study was also limited by the availability of data for inclusion in the NMA, which in some cases included relatively small sample sizes and sparse networks of evidence. This resulted in wide CIs which in some cases covered the decision threshold, raising the possibility that potential differences between interventions were not identified.

Conclusion

The findings of this SLR and NMA suggest that UMEC/VI provides better outcomes in terms of lung function (as measured by change from baseline in trough FEV1) with a similar safety profile compared with LAMA and LABA monotherapies and other LAMA/LABA dual therapies. UMEC/VI also provided additional benefits on HRQoL, symptoms, rescue medication use and moderate/severe exacerbations compared with some monotherapies. These results suggest that treatment with UMEC/VI dual therapy may improve outcomes for symptomatic patients with COPD.