Introduction
Some patients with chronic obstructive pulmonary disease (COPD) using long-acting muscarinic antagonists (LAMAs) or long-acting β
2-agonists (LABAs) also use frequent short-acting β
2-agonist (SABA) rescue therapy [
1]. An analysis of more than 23,000 patients from 23 clinical trials of mono- and dual- bronchodilators demonstrated that patients across a variety of COPD severities used approximately 4 SABA puffs/day [
2]. Patients may use high levels of SABA for a number of reasons, including having poorly controlled COPD due to suboptimal prescribing of maintenance therapy, having a mild exacerbation, poor adherence or responsiveness to maintenance therapy, or a lack of access to appropriate therapy. SABA use tends to increase with increasing COPD severity; in patients who were receiving a single LAMA or LABA bronchodilator in routine US clinical practice, a mean SABA use of 3.3 puffs/day was reported in patients with less severe airflow limitation (≥50% predicted forced expiratory volume in 1 s [FEV
1]), compared with 5.2 puffs/day in patients with more severe airflow limitation (< 50% predicted FEV
1) [
3]. This suggests that patients with COPD who frequently use SABA may be inadequately treated with their current maintenance therapies [
4].
High supplementary SABA use is a marker of an increased risk of exacerbations and hospitalisation and is associated with significant economic costs [
5,
6]. In COPD clinical trials, reductions in daily SABA use (puffs/day) have also been shown to positively correlate with mean improvements in lung function, exacerbation rates and health-related quality of life [
7]. Therefore, the assessment of rescue medication use in patients with COPD is likely to be a useful measure of changes in symptom burden in clinical trials and routine clinical practice [
7].
Although rescue medication use may be a useful indication of symptom severity there is also some evidence to suggest that high levels of rescue medication use may affect the assessment of patient-perceived symptom severity. A previous post hoc analysis of two large, 6-month bronchodilator trials suggested that symptomatic patients with high SABA use (≥3.6 puffs/day) perceive a lower benefit of treatment differences between maintenance bronchodilator therapy and placebo on dyspnoea (measured using the transition dyspnoea index [TDI]) compared with low SABA users (< 3.6 puffs/day) [
1]. This finding may have important implications for the design of clinical trials; therefore, this prespecified analysis was performed to investigate the potential confounding effect of SABA use on treatment differences observed between a LAMA/LABA combination and LAMA or LABA monotherapy in more detail.
The Early MAXimisation of bronchodilation for improving COPD stability (EMAX) trial examined the benefits of dual bronchodilation with the LAMA/LABA combination of umeclidinium/vilanterol (UMEC/VI) versus the LAMA UMEC and the LABA salmeterol (SAL) in symptomatic patients at low exacerbation risk who were not receiving inhaled corticosteroids (ICS) [
8]. The primary analysis demonstrated consistent incremental benefits of UMEC/VI compared with both monotherapies on lung function and symptoms [
8]. The objective of this pre-specified analysis of the EMAX trial was to prospectively investigate whether differences in SABA use at baseline are associated with differences in lung function and symptomatic treatment responses to dual- versus mono-bronchodilation in symptomatic patients with COPD. As such, subgroup analyses of high and low SABA users were conducted to compare treatment differences in lung function and symptoms for each subgroup.
Methods
Study design and treatments
This was a pre-specified analysis of the multicentre, randomised, double-blind, double-dummy, 3-arm parallel group EMAX trial (NCT03034915; GSK study number 201749). Full details of the study methodology have been published previously [
8]. Briefly, patients were randomised 1:1:1 to once-daily UMEC/VI 62.5/25 μg delivered via the ELLIPTA inhaler, once-daily UMEC 62.5 μg via the ELLIPTA inhaler, and twice-daily SAL 50 μg via the DISKUS inhaler, for 24 weeks.
Patients
Patients were ≥ 40 years of age with a current diagnosis of COPD according to the American Thoracic Society/European Respiratory Society definition [
9], were current or former smokers with ≥10 pack-years of smoking history, had a pre- and post-salbutamol FEV
1/forced vital capacity ratio < 0.7 and post-salbutamol FEV
1 of ≥30–≤80% of predicted (Global Initiative for COPD [GOLD] grade 2/3 lung function), a COPD Assessment Test (CAT) score ≥ 10, had ≤1 moderate exacerbation and no severe exacerbations in the prior year and were ICS free for ≥6 weeks and LAMA/LABA free for ≥2 weeks prior to 4-week run-in period. During the run-in period, patients were limited to a maximum of one bronchodilator maintenance therapy with either a LAMA or LABA. As-needed salbutamol was permitted throughout all study phases but was not permitted within the 4 h prior to spirometry testing.
Endpoints and assessments
Endpoints assessed in this pre-specified analysis included change from baseline in FEV1 at Week 24, self-administered computerised-TDI (SAC-TDI) at Week 24, change from baseline in Evaluating Respiratory Symptoms-COPD (E-RS) at Weeks 21–24, and global assessment of disease severity (GADS; change from baseline rated on a seven-point Likert scale of ‘much better’, ‘better’, ‘slightly better’, ‘no change’, ‘slightly worse’, ‘worse’, ‘much worse’) at Week 24. Daily SABA (salbutamol) use (puffs/day) was reported across the 24 weeks.
Analysis of trough FEV
1 responders, defined as patients with an improvement from baseline of ≥100 mL, was performed post hoc. SAC-TDI and E-RS responders were analysed prospectively with responders defined as patients with a ≥ 1-unit and ≥ 2-point improvement from baseline, respectively [
10,
11]. Risk of a first moderate or severe exacerbation up to Day 168 was also determined. Safety outcomes included incidence of adverse events (AEs) and serious AEs (SAEs).
To use more of the information available in the range of SABA use, fractional polynomial modelling was performed as a post hoc analysis. Changes in FEV1 and SAC-TDI at Week 24, SABA use (puffs/day) at Weeks 1–24, and E-RS at Weeks 21–24, were assessed using fractional polynomial modelling with continuous transformations of baseline SABA use as a covariate to understand if and at what level of baseline SABA use the efficacy differences between the three maintenance bronchodilator regimens were impacted.
Statistical analysis
Subgroup analyses in high and low SABA users were performed prospectively with the intent-to-treat (ITT) population of the EMAX trial stratified by median baseline SABA use (low, < 1.5 puffs/day; high, ≥1.5 puffs/day). Comparisons between baseline characteristics of low and high SABA subgroups were descriptive only.
Trough FEV1 and patient-reported outcomes were analysed using mixed model repeated measures analyses adjusted for covariates of baseline score, geographical region, number of bronchodilators per day during run-in, visit/4-weekly period, treatment, visit/4-weekly period by baseline and visit/4-weekly period by treatment interactions. Data are presented as least squares (LS) mean and LS mean change from baseline, with estimated treatment differences and 95% confidence intervals (CIs). Responder analyses with corresponding odds ratios (OR) and 95% CIs were performed using a generalised linear mixed model with covariates of baseline score, number of bronchodilators per day during run-in, geographical region, number of bronchodilators per day during run-in, visit/4-weekly period by baseline and visit/4-weekly period by treatment interactions. For the GADS, ordered ORs of a better response category (‘slightly better’, ‘better’ or ‘much better’) on the seven-point scale at each visit were analysed using a generalised linear model with covariates of treatment, geographical region and the number of bronchodilators per day during run-in. Time to study treatment withdrawal and first exacerbation hazard ratios (HR) and 95% CIs were based on Cox proportional hazards model with covariates of treatment, number of bronchodilators per day during run-in, and geographical region.
For fractional polynomial modelling, transformation of baseline SABA use were assessed using a fitted mixed model repeated measures with additional covariates of baseline endpoint value, geographical region, number of bronchodilators per day during run-in, visit/4-weekly period, treatment, first-degree fractional polynomial (FP1), FP2, visit/4-weekly period by baseline, visit/4-weekly period by treatment, FP1-treatment and FP2-treatment interactions.
Discussion
This analysis of the EMAX trial investigated whether baseline SABA use in patients with COPD is associated with the level of response to dual- versus mono-bronchodilation. Compared with UMEC/VI, a greater proportion of patients receiving UMEC or SAL withdrew from study treatment in the high SABA use subgroup. Both high and low SABA use subgroups demonstrated similar improvements in lung function with UMEC/VI versus UMEC. Patients with low baseline SABA use demonstrated significant incremental COPD symptom improvements, measured using SAC-TDI and E-RS scores, with UMEC/VI compared with UMEC. However, smaller and non-significant treatment differences were observed in the high SABA use subgroup with UMEC/VI versus UMEC for both symptom outcomes, despite a higher symptom burden in the high versus low SABA use subgroup at baseline. In contrast, the high SABA use subgroup showed significant reductions in daily rescue medication use that exceeded the non-significant reductions seen in the low SABA use subgroup; however, this may be in part due to patients in the low baseline SABA subgroup having little potential for improvement in this endpoint, since the majority were not using SABA at baseline. The findings with UMEC/VI versus UMEC might suggest that treatment differences may be impacted by concurrent use of rescue medication, with diminished efficacy of the LABA component observed on subjective symptom-based patient-reported outcomes in patients with high SABA use (approximately 4 puffs/day based on the fractional polynomial analyses) at baseline, whereas the effect was not so marked with UMEC/VI versus SAL (i.e. the addition of LAMA).
There are several possible mechanisms for the observed impact of SABA use on symptom outcomes. One possibility is the development of tolerance to β
2-agonist bronchodilators or a loss of responsiveness to LABA maintenance bronchodilators [
12]. However, as the improvements in change from baseline in lung function and symptoms were greater in high versus low SABA subgroups for all maintenance regimens, it is unlikely that tolerance to β
2-agonist bronchodilators, or a loss of responsiveness to their effects, explain the diminished efficacy differences between LAMA/LABA and LAMA therapy seen in the current analyses. There could be pathophysiological differences between high and low SABA users that are unaccounted for in our analyses and that cannot be adjusted for or explained. It has previously been suggested that SABA use may be habitual [
13]; however, our data do not support this, as differences in baseline disease characteristics suggest that the high SABA use subgroup had more severe COPD (worse lung function, a higher symptom burden and increased exacerbation incidence) compared with the low SABA use subgroup, and fewer were using long-acting maintenance therapy at baseline. Furthermore, in the high SABA use subgroup, the rates of study treatment withdrawal were higher for patients receiving monotherapy compared with UMEC/VI. This is in agreement with other studies that indicate that SABA use may be considered a marker of disease severity and increased exacerbation risk [
1,
5,
7] or a marker of suboptimal care.
An intriguing hypothesis to consider is that high SABA use in addition to a LAMA therapy may blunt the patient-perceived symptom response to LAMA/LABA therapy. Patients using high levels of SABA at baseline and retaining high SABA use on LAMA monotherapy, may not be able to distinguish between the benefit of frequent SABA and daily LABA therapy, and therefore not perceive any additional change in symptom burden in response to dual therapy, despite an improvement in lung function and a reduction in SABA use. This is supported by the finding that the between treatment differences (UMEC/VI vs either monotherapy) in SAC-TDI and change from baseline in E-RS were smaller in the high SABA use subgroup compared with the low SABA use subgroup. As such, in patients with high SABA use, symptomatic treatment response may be more effectively monitored by assessing change in rescue medication use.
In the subgroup analyses, similar findings to those observed for UMEC/VI versus UMEC were also evident when comparing symptom outcomes between UMEC/VI and SAL, with symptom improvements generally more favourable for the dual therapy in the low baseline SABA use subgroup than in the high SABA use subgroup. However, UMEC/VI was found to provide significant improvements in E-RS score and exacerbation risk reduction versus SAL in both the high and low SABA use subgroups. This suggests that high SABA use may have less of an impact upon treatment differences between LAMA/LABA versus LABA than between LAMA/LABA and LAMA. Consequently, it is possible that when interpreting symptom outcomes between different classes of maintenance bronchodilators, the type of short-acting therapy, anticholinergic or β2−agonist bronchodilator therapy, and level of concurrent SABA use could increase the complexity of data interpretation.
The potential for confounding of LABA and SABA bronchodilator efficacy on perceived dyspnoea and exacerbation outcomes with increased baseline SABA use is consistent with a previous post hoc analysis by Naya et al of two large randomised controlled trials [
1]. In the post hoc analysis, the levels of β
2−agonist rescue medication use were higher than in the EMAX trial and almost half the population were using concurrent ICS, whereas ICS use was not permitted in the EMAX trial [
1]; however, the findings presented here from a prospective analysis of the EMAX trial are generally supportive of the findings by Naya et al [
1]. Together these data suggest the need for consideration of SABA use when assessing the incremental symptom benefits observed with dual- versus mono-bronchodilator therapy.
Assessing outcomes using fractional polynomial modelling allows modelling of a non-linear relationship. This provides a wider assessment of treatment differences between maintenance therapies across covariate values than may be observed compared with pre-determined subgroups. The use of fractional polynomial analysis in this study provides more informative results than the subgroup analysis alone and indicates that the need to consider SABA use is of particular importance to patients who use high levels of SABA (> 4 puffs/day). Similar approaches have recently been used to compare the efficacy of triple ICS/LAMA/LABA therapy versus LAMA/LABA according to baseline blood eosinophil count [
14], and to evaluate the exacerbation risk reduction efficacy of budesonide–formoterol versus formoterol according to baseline blood eosinophil count [
15]. Analysing SABA use with this modelling approach may be more beneficial for physicians to determine how to optimise treatment for their patients. This may be particularly important in maintenance-naïve patients with symptomatic COPD receiving short-acting bronchodilators alone, who may have developed a reliance on rescue bronchodilator therapy.
The following limitations should be considered when interpreting the results of these analyses. As there is currently no consensus for what constitutes low or high SABA use among patients with COPD, and level of use was unknown pre-study, we pre-specified the median SABA use (1.5 puffs/day) as the subgroup threshold to generate numerically balanced subgroups. The high and low SABA subgroups were nonetheless large and baseline clinical characteristics were well balanced between treatment arms within each subgroup. Also, as approximately a third of patients were not using SABA at baseline, an analysis comparing three subgroups of no, low and high SABA users may be of interest; however, this post hoc analysis would result in small, potentially less well-balanced subgroups and would provide little additional information than that already provided by the fractional polynomial analyses, which include estimates of efficacy differences for non-SABA users. The EMAX population had a low risk of exacerbation; as a consequence of this and the relatively short study duration, the subgroup analyses were not powered to detect exacerbation treatment differences, and so assessment of the impact of high versus low SABA use on the risk of exacerbation was limited. As VI is not available as a licensed drug, SAL was used as the comparator LABA in this study. However, evidence suggests that the efficacy of VI and SAL are similar on trough FEV
1 and TDI [
16]. Finally, the fractional polynomial analysis suggests that most of the diminished treatment differences for symptom-based outcomes between the LAMA/LABA and LAMA monotherapy were observed in the 18% of EMAX patients who used ≥4 puffs/day of SABA; consequently our findings should not be generalised to populations with less frequent SABA use.
Competing interests
IHB, PWJ, CC and DAL are employees of GlaxoSmithKline (GSK) and hold stocks and shares in GSK. IPN was an employee of GSK at the time of the study, holds stocks and shares in GSK and is a contingent worker on assignment at AstraZeneca. LT is a contingent worker on assignment at GSK. FM has received research grants for participating in multicentre trials for AstraZeneca, Boehringer Ingelheim, GSK, Sanofi and Novartis, and has received unrestricted research grants and personal fees from Boehringer Ingelheim, Grifols and Novartis. CFV has received grants from AstraZeneca, Boehringer Ingelheim, Chiesi, GSK, Grifols, Mundipharma, Novartis and the German Federal Ministry of Education and Research (BMBF) Competence Network Asthma and COPD (ASCONET), and has received personal fees from AstraZeneca, Boehringer Ingelheim, Berlin Chemie/Menarini, Chiesi, CSL Behring, GSK, Grifols, MedUpdate, Mundipharma, Novartis, Nuvaira and Teva. LB has received honoraria for giving a lecture or attending an advisory board for Airsonett, ALK-Abello, AstraZeneca, Boehringer Ingelheim, Chiesi, GSK, Meda, Novartis and Teva. EMK has served on advisory boards, speaker panels or received travel reimbursement from Amphastar, AstraZeneca, Boehringer Ingelheim, GSK, Mylan, Novartis, Pearl, Sunovion, Teva and Theravance, and has received consulting fees from Cipla and GSK. ELLIPTA and DISKUS are owned by/licensed to the GSK group of companies.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.