Long-acting bronchodilators improve exercise capacity in COPD patients: a systematic review and meta-analysis

Background
We carried out a systematic review and meta-analysis with the aim to evaluate the efficacy of longacting
bronchodilators on exercise capacity in COPD patients.


Methods
The endpoints were the efficacy of long-acting bronchodilators (altogether, and by single classes)
vs. placebo in modifying endurance time (ET), inspiratory capacity (IC) and dyspnea during exercise, taking into consideration the outcomes according to different patients’ inclusion criteria and exercise methodology.


Results
Twenty-two studies were deemed eligible for analysis. Weighted mean increase in ET resulted of 67 s
(95% CI ranges from 55 to 79). For isotime IC and dyspnea during exercise, weighted improvements were 195 ml (162–229), and − 0.41 units (− 0.56 to − 0.27), respectively. The increase in trough IC was 157 ml (138–175). We found a trend in favour of LAMA compared to LABA in terms of ET. In the 11 studies which reported a value of functional residual capacity > 120% as inclusion criterion, weighted mean increase in endurance time was 94 s (65 to 123); however we did not find any significant correlation between ET and mean trough IC (P: 0.593). The improvement of ET in the 5 studies using walking as exercise methodology resulted of 58 s (− 4 to 121).


Conclusions
Long-acting bronchodilators improve exercise capacity in COPD. The main effect of long-acting bronchodilators seems to be a increase of basal IC rather than a modification of dynamic hyperinflation during exercise. The efficacy in terms of endurance time seems higher in studies which enrolled patients with hyperinflation, with a similar efficacy on walking or cycling.


Background
In patients with chronic obstructive pulmonary disease (COPD) exercise limitation is mainly due to dynamic hyperinflation [1], even if the contribution of other factors, such as an imbalance between respiratory and locomotor muscles for limited energy supply [2], limb muscle dysfunction [3], and co-morbidities (e.g., left ventricle diastolic dysfunction) [4] can play a significant role. Reduced daily activity has been well documented in COPD patients, resulting from both respiratory and non-respiratory clinical conditions associated with the disease [5]. An evaluation performed by multisensory armband confirmed that daily physical activity is mainly associated with dynamic hyperinflation, regardless of COPD severity [6]. It was clearly proved that exercise capacity and daily activity are closely associated with life expectancy [7]. Thus, improving physical activity represents the best approach to address both pulmonary and systemic manifestations of the disease [8].
Endurance time during high-intensity constant-load ergometer exercise protocols (i.e. 75-80% peak work-rate) is currently used to assess the effects of pharmacological and non-pharmacological interventions, as it has proven to be more sensitive than other procedures [9]. Importantly, endurance test is also listed as a suitable outcome for pivotal trials with pharmacological interventions by regulatory bodies, e.g. the Food and Drug Administration (FDA) and the European Medicines Agency (EMA), although no label claim for improvements in exercise capacity has been granted this far [10,11]. Moreover, this approach allows an evaluation of symptoms intensity (e.g., dyspnea and leg effort), or physiological variables (e.g., inspiratory capacity, IC, assessment of dynamic hyperinflation) at a standardized time (isotime), which has been proved very useful in identifying the underlying physiological mechanisms responsible for modifications in exercise tolerance induced by a particular intervention [12].
In COPD exercise capacity can be improved by rehabilitative interventions [13], as well as interventions aimed at unloading the respiratory system, such as breathing heliox (i.e.) [14], oxygen therapy [15], or noninvasive ventilatory support [16], and recently high flow nasal cannula [17]. However, the first-line treatment for all COPD patients, after smoking cessation, remains pharmacological, mainly with inhaled medication. Longacting bronchodilators represent the cornerstone of COPD pharmacological therapy [18,19]. Long-acting muscarinic antagonists (LAMA), β 2 -agonists (LABA), alone or in combination with inhaled steroids (ICS), and the recent LABA/LAMA fixed-dose combinations (FDCs), have been evaluated to assess their efficacy on exercise capacity [8]. Despite the availability of studies on LABA, LAMA, and ICS/LABA or LABA/LAMA FDCs on exercise capacity, evidence for the efficacy of bronchodilators in enhancing the exercise capacity of patients with COPD is sometimes contradictory [20], with, for instance, a recent trial which failed to demonstrate the superiority of a LABA/LAMA FDC vs. placebo [21]; moreover, current literature is characterized by a large heterogeneity of studies, due to differences in terms of inclusion criteria (unselected patients vs. COPD patients with evidence of hyperinflation), and/or exercise methodology (cycling vs. walking, cycle or treadmill ergometer vs. shuttle walking test).
On this basis, we carried out a systematic review and meta-analysis on the available clinical evidence to evaluate the efficacy of long-acting bronchodilators (altogether, and by single classes) on exercise capacity, dynamic hyperinflation, and dyspnea during exercise using high-intensity constant-load exercise test both in unselected patients and in patients with demonstrated lung hyperinflation at rest, and in studies with different exercise methodologies (walking vs. cycling).

Methods
This systematic review and meta-analysis was conducted according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) statement [22].

Search strategy
We selected randomized controlled studies (RCTs) focused on the efficacy of long-acting bronchodilators (i.e.: LABA, LAMA, LABA/ICS, and LABA/LAMA FDCs) on exercise capacity in COPD patients based on endurance time with high-intensity constant-load performed on a cycle or treadmill and whose duration was longer than one week. We searched in PubMed and EMBASE through September 2017. The following keywords and their related MeSH (Medical Subjects Heading) terms were used: "chronic obstructive pulmonary disease", "COPD", "bronchodilator", "long-acting bronchodilator", "LABAs", "LAMAs", "cardiopulmonary exercise test", "endurance time", and "exercise". Only publications in English language were considered.

Studies selection and data extraction and study quality assessment
We included RCTs focused on COPD patients (any level of severity) which assessed as a primary or secondary outcome the efficacy of long-acting bronchodilators on exercise capacity evaluated by "endurance time" using high-intensity constant-load ergometer exercise test, or shuttle walking test, providing the results of active drugs and placebo and not only between-arms difference. The shuttle walking test uses an audio signal from a tape cassette to direct the walking pace of the patient back and forth on a 10-m course. Test ends when the patient cannot reach the turnaround point within the required time [12]. We included studies which used cycling or walking. In both cases maximal exercise capacity was measured using incremental symptom limited exercise test. Subsequently, constant-load (cycling or walking at 75-90% of maximal work load) cycling ergometer or shuttle walking test (ESWT) were used to compare the efficacy of active drugs with placebo. The following exclusion criteria were chosen: 1) manuscripts focused on shortacting bronchodilators, or not including placebo; 2) exercise capacity assessment based on a protocol different from high-intensity constant load exercise test; 3) epidemiological observational study design or experimental design other than RCT; 4) manuscripts not written in English language; 5) data expressed only as difference vs. placebo and not as single arms (active drugs and placebo). Two independent authors (MAR and FDM) firstly reviewed all titles/abstracts to identify potentially relevant articles. Then, study selection, based on a full-text review, was performed according to the predefined inclusion/exclusion criteria and disagreements were resolved by consensus. The same authors reviewed eligible studies using the CONSORT quality standard, judged the studies quality by Jadad scale [23,24], and the risk of bias by a domain-based evaluation, which included the following domains: sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, attrition, selective reporting bias and other sources of bias. Explicit judgments were made about the overall risk of bias according to the Cochrane guidance [25].

Endpoints
The primary endpoint was to assess the efficacy of all long-acting bronchodilators (i.e.: LABA, LAMA, LABA/ ICS FDCs, and LABA/LAMA FDCs) vs. placebo on endurance time. The secondary endpoints were to investigate the role of the above-mentioned drugs on IC and dyspnea, as well as the comparison between drug classes in terms of endurance time, IC, and dyspnea.

Statistical analysis
Forest plots were created to graphically assess both the variability of the sample estimates and the weight of sample sizes in the computation of estimates (weighted means). A random-effects meta-analysis was carried out to account for the presumed high between-study variability. Inconsistency among studies was assessed by the I 2 statistic to underscore the effect of true variability on the overall variation. Since the inclusion criteria of the studies in this field sometimes include the presence of hyperinflation (i.e. FRC > 120%) a subgroup analysis was carried out; furthermore, sub-analyses of the endurance time were performed according to the exercise methodology (i.e. cycling vs. walking). Funnel plots and their related Egger's tests were performed to visually assess the risk of bias, particularly publication and small sample bias. Correlation between endurance time and change in relevant physiological variables was undertaken. The statistical software used were Stata13.0 (StataCorp, College Station, TX, USA) and StatsDirect 2.8.0, version 1.4 (StatsDirect Ltd., Altrincham, UK).

Results
Out of 88 potentially relevant studies, 22 (25%) were deemed eligible for a qualitative and quantitative analysis (Fig. 1); their characteristics are summarized in Table 1 [21, [26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44]. The efficacy of LABA, LAMA, LABA/ LAMA FDCs, and ICS/LABA FDCs as main treatment was assessed in 6 (27%), 10 (45%), 3 (14%), and 3 (14%) studies, respectively, with 5 (23%) studies using another active drug as control of the experimental arm ( Table 1). The study design was cross-over in the majority of the cases (15 studies, 68%). Endurance time was the most frequently adopted primary outcome, followed by pulmonary function; mean duration of treatment was 8.9 ± 15.9 weeks. The methodology used for exercise and the intensity of the constant workload are reported in Table 1; in most cases cycle ergometer exercise testing was performed at 75-90% of maximal work load. The BORG scale was used to evaluate dyspnea in all studies. Characteristics of the enrolled patients are reported in Table 2. Total sample size included 2898 patients; 65.4 ± 9.1% of which were males, with a mean age (mean SD)   The median Jadad score for the RCTs included in our analysis was 4 (range 4-5); a detailed assessment of the risk of bias is described in the supplement materials using the Cochrane collaboration tool for assessing risk of bias. The risk of bias was deemed low for the majority of the items in the selected studies. No relevant asymmetries were found in several funnel plots related to the main clinical variables.
Efficacy of long-acting bronchodilators on ET, inspiratory capacity, and dyspnea during exercise Efficacy of different classes of long-acting bronchodilators Figure 5 shows the efficacy of different classes of longacting bronchodilators at approved doses for COPD treatment in terms of endurance time, isotime IC, and dyspnea. This analysis did not show any significant differences among the investigated categories of longacting bronchodilators; however, for endurance time the efficacy of LABA seems lower than for other classes.

Efficacy of long-acting bronchodilators in patients with hyperinflation and using different exercise methodologies
To better understand the role of hyperinflation at rest we performed a prespecified subgroup analysis focused on the 11 studies which reported a value of FRC > 120% as inclusion criterion (Table 2). In this subgroup of studies, mean change of trough FEV 1 following the exposure to all long-acting bronchodilators (end of the study) resulted of 153 ml (95% CI ranges from 123 to 183; I 2 : 73.8% treatment arms: 19), and mean change of trough IC resulted of 154 ml (95% CI ranges from 121 to 187; I 2 : 50.4% treatment arms: 18). Weighted mean change of endurance time, inspiratory capacity, and dyspnea following the exposure to long-acting bronchodilators resulted of 94 s (95% CI ranges from 65 to 123; I 2 : 10.4%, computed on 11 treatment arms), 174 ml (95% CI ranges from 131 to 216; I 2 : 0%; treatment arms: 14), and − 0.37 units (95% CI ranges from − 0.54 to − 0.21; I 2 : 53.2%; treatment arms: 18), respectively (Additional file 1: Figure S2A). In the subgroup of the 11 studies which did not require an increase of FRC as inclusion criterion weighted mean change of endurance time, inspiratory capacity and dyspnea following the exposure to long-acting bronchodilators resulted of 61 s (95% CI ranges from 49 to 73; I 2 : 17.2%, computed on 23 treatment arms), 231 ml (95% CI ranges from 178 to 285; I 2 : 0%; treatment arms: 6), and − 0.51 units (95% CI ranges from − 0.80 to − 0.21; I 2 : 60.2%; treatment arms: 12), respectively (Additional file 1: Figure S2B). We did not find any significant correlations between ET and mean trough FEV 1 (rho: 0.38, P: 0.063), or between ET and trough IC (rho: 0.14, P: 0.593).
In the subgroup of the 5 studies which used walking as exercise methodology, weighted mean change of endurance time following the exposure to long-acting bronchodilators resulted of 58 s (95% CI ranges from − 4 to 121; I 2 : 56.2%, computed on 5 treatment arms), compared to 68 s (95% CI ranges from 56 to 79; I 2 : 13.3%, computed on 29 treatment arms) in the studies which used a cycle ergometry (Additional file 1: Figure S3).

Discussion
The main findings of this systematic review and metaanalysis are: 1) long-acting bronchodilators are effective in improving exercise capacity in COPD patients, with an average increase of about 60 s; 2) this effect is Fig. 3 Efficacy of long-acting bronchodilators on isotime inspiratory capacity. I 2 1.2% (95% CI 0-42.8%). Error bars represent 95% confidence intervals. *: study n. 417 associated with a reduction of dyspnea during exercise; 3) the increase in isotime IC is similar to the change of trough IC, and this is the reason why the effect seems to be a decrease of basal inspiratory capacity rather than a modification of dynamic hyperinflation during exercise; 4) the use of LABA and LAMA is associated with a similar improvement of exercise tolerance, dynamic hyperinflation, and dyspnea; however, it is possible to identify a trend in favour of LAMA in terms of ET; 5) the efficacy of long-acting bronchodilators on ET were higher in the studies which required an increase of FRC as inclusion criterion; however we failed to demonstrate a correlation between ET and trough IC; 6) the efficacy of long-acting bronchodilators is similar when walking or cycling are used as exercise methods.
Maintenance therapy with long-acting bronchodilators has been demonstrated to significantly reduce operational lung volumes during exercise in symptomatic patients with COPD, even if this beneficial effect does not always have an impact on exercise duration. Casaburi R et al. found a significant effect of tiotropium on exercise duration in COPD patients with moderate (i.e. FEV 1 < 80%), but not mild (i.e. FEV 1 > 80%) disease [40]. Our meta-analysis found that, on average, there is a concordance between the increase of inspiratory capacity, the improvement of exercise tolerance, and the reduction of dyspnea in patients with moderate to severe COPD. Unfortunately, we are not able to extend these findings to patients with mild disease, since most of the studies we included in our analysis had enrolled patients with FEV 1 < 70% predicted value (Table 2), and also in Fig. 4 Efficacy of long-acting bronchodilators on dyspnea. I 2 55.1% (95% CI 26.8-69.4%). Error bars represent 95% confidence intervals. For Ume/ Vil †: study n. 1222.37; *: study n. 1222.38. For Olo †: study n. 418; *: study n. 417 case of inclusion criteria permissive for milder disease, results for patients with FEV 1 > 80% were not available. As previously discussed, we found a significantly higher improvement in ET, of about 50%, in the studies which required an increase of FRC as inclusion criterion; this result could demonstrate that long-acting bronchodilators are more effective in terms of exercise capacity in COPD patients with lung hyperinflation. Thus, even if we found that the improvement of operational lung volume is correlated with the increase of ET after bronchodilation, we failed to demonstrate a correlation between trough IC and ET per se, result which could suggest that, even if hyperinflation is a crucial aspect for exercise limitation in COPD, the absolute value of ET depends on many factors.
We did not find any significant difference in terms of endurance time, dynamic hyperinflation, and symptoms during exercise between LABA and LAMA, the two classes of long-acting bronchodilators most studied in these fields; however, a trend in favor of LAMA was identifiable at least in terms of endurance time. It's to notice that the present study does not allow to explain why LAMA could be superior of LABA in terms of exercise capacity. Beta-2 agonists and antimuscarinic agents could theoretically exert a different efficacy on exercise capacity due to differences in the distribution of receptors, both in and outside the lung, and in mechanism of action. Indeed, beta-2 adrenergic receptors are present in high density not only in airway smooth muscle cells, but also in vascular endothelium, ciliated cells, circulating inflammatory cells, such as eosinophils, and sub-mucosal glands. On the other hand, muscarinic receptors are present on bronchial smooth muscle cells, parasympathetic and sympathetic nerves, ganglia cells, mucus secreting cells, and inflammatory cells. ICS/LABA FDCs have proved to be more effective than LABA alone in terms of function, Fig. 5 Efficacy of different classes of long-acting bronchodilators tested at the approved dose for COPD treatment. Error bars represent 95% confidence intervals quality of life, symptoms and prevention of exacerbations [45]. Therefore, even if the efficacy on FEV 1 cannot be directly translated to lung deflation (a crucial determinant of exercise capacity on COPD), there is a rationale to expect a better efficacy of ICS/LABA FDCs rather than LABA alone on exercise capacity. The same, and probably with major emphasis, has to be expected for LABA/LAMA combinations, which confirmed to be more effective in terms of function and symptoms than not only LABA or LAMA alone, but also ISC/LABA FDCs [46,47]. Even if some of the studies currently available failed to demonstrate a superiority of LABA/LAMA compared with monocomponents [21,41], in a recent meta-analysis LABA/ LAMA FDCs were found to be more effective than LABA or LAMA alone in terms of ET [48]. There is limited information on minimum clinically significant differences for endurance time after an intervention [49]. In COPD, a 105 s change from baseline using cycle ergometry related well with positive patient-reported outcomes after pulmonary rehabilitation [50,51]. However, data from bronchodilator studies suggest that improvements in lung function that seem to be clinically important are often associated with increases in endurance time of 60 s [49]. For dyspnea (isotime Borg) and IC at isotime, the minimal clinically important difference (MCID) were suggested to be 1 unit, and 200 ml, respectively [49]. Our meta-analysis found a weighted mean change, following the exposure to long-acting bronchodilators, of endurance time, dyspnea, and IC at isotime of 67 s, − 0.41 units, and 195 ml, respectively. These improvements we found are higher than the MCID for endurance time, probably the most important outcome, borderline clinically significant for dynamic hyperinflation during exercise, but far from the MCID for dyspnea. The average improvement we found is similar to that obtained by oxygen in mildly or nonhypoxaemic COPD patients who are dyspnoeic at rest [52]. It is noteworthy that, as pointed out by Puente-Maestu et al. in a recent ERS statement, nonpharmacological interventions, such as rehabilitation, Heliox, oxygen in hypoxaemic patients and noninvasive ventilation, have demonstrated significantly higher endurance time increases [49]. However, pharmacological and nonpharmacological treatments of COPD are not mutually exclusive, since the former have to be considered in all cases, and their combination has shown to be more convenient that a single approach [33]. Moreover, the evaluation of variables at isotime is sometimes problematic since they are often extrapolated or interpolated.
Finally, we found a similar efficacy of long-acting bronchodilators when walking or cycling are used as exercise methodologies. The main reason why the sub-analysis of the 5 studies which used walking as exercise methodology failed to demonstrate a statistically significant efficacy vs. placebo is probably due to the low power of the analysis, with an average difference similar to that of the studies which used cycle ergometer (58 vs. 68 s). This result, important since "walking" is evidently the most common real-life exercise of COPD patients, confirms a previous study, which demonstrated a similar efficacy of formoterol on endurance time evaluated by walking and cycling [53].
A number of potential limitations of the study deserves discussion. Firstly, the studies we included in this metaanalysis present different inclusion criteria for COPD patients who, in most but not all studies, were requested to present a significant hyperinflation (e.g. FRC > 120% of predicted values); this could limit the external validity of our analysis, since the main indication of long-acting bronchodilators is so far COPD per se (i.e. FEV 1 /FVC < lower limit of normal) without the need of hyperinflation. However, a specific analysis including the studies which requested an increase of FRC as inclusion criterion was performed, with evidence of a higher efficacy of longacting bronchodilators when compared with the studies which evaluated unselected COPD patients. Secondly, the duration of the studies we included in our analysis was very heterogeneous, ranging from one to 96 weeks, with an average duration of the treatment of 9.6 weeks. Probably depending on the onset of action, some drugs demonstrated to be able to improve endurance time immediately after the first inhalation [29]; however, due to the chronic nature of the disease and the consequent long treatment course, this aspect is not so clinically relevant, and a significant effect expected in "some" weeks is the message which emerges from our analysis. Then, we decided to include ICS/LABA combination in the analysis, since this treatment is very common in clinical practice. The addition of ICS to LABA can be a confounder for the interpretation of the effect of long-acting bronchodilators on exercise capacity. However, the subgroups analysis according to treatment classes (LABA; LAMA, ICS/LABA, and LABA/LAMA) allows to evaluate the effects of bronchodilators alone, confirming the efficacy of these treatment without ICS. Finally, the changes we found of isotime IC are close to those of trough IC, suggesting that the main effect of long-acting bronchodilators is an increase in basal inspiratory capacity rather than a modification of dynamic hyperinflation during exercise (i.e. the slope of change of IC).

Conclusion
Long-acting bronchodilators improve exercise capacity in COPD patients, with an average change higher than the minimum clinically significant difference. While long-acting bronchodilators consistently improve the capacity to tolerate exercise by improving lung mechanics, the challenge remains to convert this advantage into increased habitual physical activity. An additional behavioral intervention will likely be needed to achieve this in the hope of improving long-term survival [8].

Additional file
Additional file 1: Figure S1. Funnel plots for ET, isotime IC and isotime dyspnea. Figure S2A Forest and Funnel plots of ET, isotime IC and isotime dyspnea in the 11 studies which included only COPD patients with functional residual capacity (FRC) > 120%. Figure S2B Forest and Funnel plots of ET, isotime IC and isotime dyspnea in the 11 studies which did not require an increase of functional residual capacity (FRC) as inclusion criterion (i.e. unselected COPD patients). Figure S3 Forest and Funnel plots of ET of the 5 studies (A) which used walking, and 17 studies (B) which used cycling as exercise methodology to assess the efficacy of long-acting bronchodilators. Table S1 Cochrane