LABA/LAMA combination in COPD: a meta-analysis on the duration of treatment

Abstract

When there are no randomised clinical trials directly comparing all relevant treatment options, an indirect treatment comparison via meta-analysis of the available clinical evidence is an acceptable alternative. However, meta-analyses may be very misleading if not adequately performed. Here, we propose and validate a simple and effective approach to meta-analysis for exploring the effectiveness of long-acting β₂-agonist (LABA)/long-acting muscarinic antagonist (LAMA) fixed-dose combinations in chronic obstructive pulmonary disease.

14 articles with 20 329 patients (combinations n=9292; monocomponents n=11 037) were included in this study. LABA/LAMA combinations were always more effective than the monocomponents in terms of the improvement in trough forced expiratory volume in 1 s, transition dyspnoea index and St George's Respiratory Questionnaire scores after 3, 6 and 12 months of treatment. No significant publication bias was identified. Significant discrepancies with previous network meta-analyses have been found, with overall differences ranging from 26.7% to 43.3%.

Results from previous network meta-analyses were misleading because no adequate attention was given to formulating the review question, specifying eligibility criteria, correctly identifying studies, collecting appropriate information and deciding what it would be pharmacologically relevant to analyse. The real gradient of effectiveness of LABA/LAMA fixed-dose combinations remains an unmet medical need; however, it can be investigated indirectly using a high-quality meta-analytic approach.

Short abstract

We propose a simple and effective meta-analytic approach for exploring the impact of LABA/LAMA combinations in COPD http://ow.ly/8Zd9302154B

Introduction

The wide availability of long-acting β₂-agonist (LABA)/long-acting muscarinic antagonist (LAMA) fixed-dose combinations (FDCs) in the absence of head-to-head comparative randomised clinical trials (RCTs) makes difficult the choice of combination for the treatment of chronic obstructive pulmonary disease (COPD), a heterogeneous disorder characterised by an enormous variability in drug response between phenotypes [1, 2].

Some meta-analyses have been undertaken in an attempt to improve the gaps in our knowledge in this field [3–5]. Since no RCTs exist that have directly compared all relevant treatment options, an indirect treatment comparison via a meta-analytic approach can be an acceptable alternative to quantifying scientific uncertainty. However, only the meta-analysis published by Oba et al. [4] has investigated the impact of LABA/LAMA FDCs in COPD with specific regard to the duration of treatment. These authors performed a sophisticated meta-analysis [4], but we have identified several critical points in their method that have affected the results and produced quite inconsistent conclusions.

In particular, we strongly believe that comparing LABA/LAMA FDCs with placebo is anachronistic. In fact, there is solid evidence suggesting that the so-called “dual” bronchodilator therapy has an important role in optimising bronchodilation [6–8]. In any case, the effectiveness of dual bronchodilation should always be compared with at least one of the monocomponents included in the FDCs, and not with different LABAs or LAMAs [5]. We also believe that the regimen of administration should be consistent between the drugs of each combination, i.e. both medications should be administered once daily, for example, and not one on a once-daily and the other on a twice-daily basis [5].

A correct meta-analytic approach should have both quantitative and qualitative characteristics, in order to reveal the biases, strengths and weaknesses of analysed studies [9]. Although the Jadad score rates three out of 10 dimensions of RCT quality, it represents an easy, fast and objective tool for assessing the quality of RCTs, by revealing great overlap with the Cochrane Back Review Group [10]. In spite of this, Oba et al. [4] selected the studies to be included in their network meta-analysis according to personal opinion [4]. Moreover, they omitted to examine several important outcomes after 12 months of treatment, although RCTs in which these variables were available at week 52 were included in their analysis [4].

More surprising is that the flow diagram for the identification of the included studies was not reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [11].

We have recently performed an indirect treatment comparison of the currently approved LABA/LAMA FDCs in COPD [5], in which the variables have been meta-analysed regardless of the duration of the treatment. The aim of the present meta-analysis was to explore whether the duration of treatment could in some way influence the clinical effectiveness of LABA/LAMA FDCs. Therefore, we have now adequately assessed the impact of the duration of treatment with LABA/LAMA FDCs on the trough forced expiratory volume in 1 s (FEV₁), St George's Respiratory Questionnaire (SGRQ) and transition dyspnoea index (TDI) in COPD, by extracting each variable at the end of RCTs lasting ≥3 months and performing a high-quality meta-analysis.

Materials and methods

Search strategy

This systematic review was performed in agreement with the PRISMA statement (supplementary table S1) [12].

We performed a comprehensive literature search for RCTs written in English regarding the impact of treatment with LABAs and LAMAs administered in combination in patients suffering from COPD and lasting ≥3 months [13, 14].

The terms “chronic obstructive pulmonary disease” and “COPD” were searched for the disease, the terms “LABAs” and “LAMAs” for the class of drugs, the terms “aclidinium”, “formoterol”, “glycopyrronium”, “indacaterol”, “olodaterol”, “tiotropium”, “umeclidinium” and “vilanterol” for specific compounds and the term “combination” to identify RCTs investigating combination therapy. The search was performed on PubMed and Google Scholar in order to provide for relevant studies published up to April 15, 2016 [15]. Further searches were performed on www.clinicaltrials.gov, the EU Clinical Trials Register and the 2015 European Respiratory Society (ERS) International Congress abstracts in order to find RCTs not yet published. Citations of previous published meta-analyses and relevant reviews were checked to select further pertinent studies, if any [16–19].

Study selection

RCTs involving COPD patients that had received inhalant administration of LABA/LAMA combinations versus at least one monocomponent were included in the analysis. Two reviewers independently examined the RCTs found from literature and databases, and any difference in opinion about eligibility was resolved by consensus.

Quality score and risk of bias assessment

The Jadad score, with a scale of 1–5 (score of 5 being the best quality), was used to assess the quality of the papers concerning the likelihood of bias related with randomisation, double blinding, withdrawals and dropouts [20]. Two reviewers independently assessed the quality of individual studies, and any difference in opinion about the quality score was resolved by consensus. RCTs with Jadad score ≥3 were included in the meta-analysis.

The risk of publication bias was analysed by applying Egger's test in order to assess funnel plot symmetry through the following regression equation: SND = a + b × precision, where SND represents the standard normal deviate (treatment effect divided by its standard error), and precision represents the reciprocal of the standard error [21–24]. Evidence of asymmetry from Egger's test was considered to be significant for p<0.1, and the graphical representation of 90% confidence bands have been presented [21–24].

Data synthesis and analysis

The end-point of this meta-analysis was to compare the impact of LABA/LAMA combinations with respective monocomponents, with regard to changes from baseline in trough FEV₁, SGRQ and TDI score at specific time points (3, 6 and 12 months).

We performed a pair-wise meta-analysis and a sensitivity analysis via a network meta-analytic approach in order to validate the obtained results. Since the data were selected from a series of studies performed by researchers operating independently, and a common effect size cannot be assumed, we used the random-effects model to perform the pair-wise meta-analysis in order to balance the study weights and to adequately estimate the confidence interval of the mean distribution of drugs effect on the investigated variables [25–28]. Values are presented as mean difference and 95% CI. Moderate to high levels of heterogeneity were considered for I²>50% [29].

The network meta-analysis was performed using a full Bayesian evidence network (chains: 4; initial values scaling: 2.5; tuning iterations: 20 000; simulation iterations: 50 000; and tuning interval: 10), the convergence diagnostics for consistency and inconsistency was assessed using the Brooks–Gelman–Rubin method [30, 31]. Results of the network meta-analysis are presented as mean difference and 95% credible level (Crl). Due to the complex evidence structure, the inconsistency of evidence has been assessed in addition to heterogeneity obtained from the pair-wise meta-analysis. While heterogeneity represents between-study variation in the measured relative effect of a pair of treatments, inconsistency can only occur when one of the treatments has a different effect when it is compared with the others [31]. The probability that each intervention arm was the most effective was calculated by counting the proportion of iterations of the chain in which each intervention arm had the highest mean difference, and the surface under the cumulative ranking (SUCRA), which represents the summary of these probabilities, was also calculated. The SUCRA is 100% when a treatment is certain to be the best, and 0% when a treatment is certain to be the worst [32].

The percentage of differences between the summary effects resulting from network meta-analysis and pair-wise meta-analysis was calculated in order to compare the results obtained from these different meta-analytic approaches. In particular, Gini's mean absolute difference (GMD) between each observation and any other observation has been calculated as previously described [33]. The GMD is defined in terms of absolute deviations averaging differences between pairs of observations, as defined by the following equation: GMD = ∑|X_i−X_j|/n [33].

The optimal information size was calculated as previously reported [34, 35], and the statistical significance was assessed for p<0.05 [36, 37].

OpenMetaAnalyst [36] was used to perform the pair-wise meta-analysis, GeMTC [38] for network meta-analysis and GraphPad Prism (La Jolla, CA, USA) software to chart the data.

Results

Study characteristics

In accordance with our hypothesis, and by applying the Jadad score for excluding low-quality studies (cut-off ≥3), we identified seven RCTs (table 1) to be excluded from the meta-analysis of Oba et al. [4], and one recent study [46] to be included. The PRISMA flow diagram and the selected RCTs are reported in supplementary figure S1 and table 2, respectively.

TABLE 1.

Reasons of exclusion of some of the studies included into the network meta-analysis of Oba et al. [4]

First author [ref.]	Reasons for exclusion
Aaron [ 39 ]	Inconsistency of the regimen of administration (tiotropium once daily versus salmeterol twice daily)
Dahl [ 40 ]	Comparison versus placebo (indacaterol/glycopyrronium versus placebo)
Maleki-Yazdi [ 41 ]	Comparison versus a monocomponent not included in the combination (umeclidinium/vilanterol versus tiotropium)
Tashkin [ 42 ]	Inconsistency of the regimen of administration (tiotropium once daily versus formoterol twice daily)
Vogelmeier [ 43 ]	Jadad score <3
Mahler [ 44 ]	Jadad score <3
Novartis Pharmaceuticals [45]	Comparison versus a monocomponent not included in the combination (indacaterol/glycopyrrolate versus tiotropium)

TABLE 2.

Patient demographics, baseline and study characteristics

Study; first author [ref.]	Study characteristics	Duration of study weeks	Patients analysed	Drugs	Regimen of administration	Patient characteristics	Age years	Male %	Current smokers %	Smoking history pack-years	Post-bronchodilator FEV1 % pred	Jadad score
NCT01604278; Vincken [ 47 ]	Multicentre, randomised, double-blind, parallel-group, placebo-controlled	12	432	Glycopyrronium/indacaterol	Once daily	Moderate to severe stable COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	63.8	48.2	51.1	44.5	54.9	4
NCT01323660, NCT01328444; Maltais [ 48 ]	Multicentre, randomised, placebo-controlled, parallel-group	12	768	Umeclidinium/vilanterol	Once daily	Moderate to severe stable COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥35% and ≤70% pred)	62.0	56.4	62.0	48.1	51.3	4
NCT01694771, NCT01696058; ZuWallack [ 49 ]	Multicentre, replicate, randomised, double-blind, parallel-group	12	2204	Tiotropium/olodaterol	Once daily	Moderate to severe COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	64.3	51.7	49.0	49.0	53.7	3
NCT01727141, NCT01712516; Mahler [ 46 ]	Identical, multicentre, randomised, double-blind, parallel-group, placebo- and active-controlled	12	1511	Glycopyrronium/indacaterol	Twice daily	Stable COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	63.5	64.8	52.2	>10	54.6	5
NCT01964352, NCT02006732; Singh [ 50 ]	Multinational, replicate, randomised, double-blind, placebo-controlled, parallel group	12	1169	Tiotropium/olodaterol	Once daily	Moderate to severe COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	64.8	61.2	47.7	>10	55.1	3
NCT01313650; Donohue [ 51 ]	Multicentre, randomised, double-blind, placebo controlled, parallel-group	24	1252	Umeclidinium/vilanterol	Once daily	COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≤70% pred)	63.3	70.7	48.2	46.0	47.6	4
NCT01313637; Celli [ 52 ]	Multicentre, randomised, placebo-controlled, parallel-group	24	1214	Umeclidinium/vilanterol	Once daily	COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≤70% pred)	63.1	65.6	50.8	44.1	48.3	4
NCT01316900, NCT01316913; Decramer [ 53 ]	Multicentre, randomised, blinded, parallel-group, double-dummy	24	1274	Umeclidinium / vilanterol	Once daily	COPD (categories B or D)	63.7	67.4	44.3	45.4	47.3	5
NCT01437397; D’Urzo [ 54 ]	Multicentre, randomised, double-blind, placebo-controlled	24	1337	Aclidinium/formoterol	Twice daily	Moderate to severe stable COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	64.1	53.3	51.7	52.5	53.7	3
NCT01462942; Singh [ 55 ]	Multicentre, randomised, double-blind, parallel-group, active- and placebo-controlled	24	1366	Aclidinium/formoterol	Twice daily	Moderate to severe COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	63.2	67.6	47.3	>10	54.3	4
NCT01202188; Bateman [ 56 ]	Multicentre, randomised, double-blind, parallel-group, placebo- and active-controlled	26	1423	Glycopyrronium/indacaterol	Once daily	Moderate to severe stable COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥30% and <80% pred)	64.0	77.2	40.0	>10	55.2	4
NCT01431274, NCT01431287; Buhl [ 57 ]	Multicentre, multinational, replicate, randomised, double-blind, active-controlled, five-arm, parallel-group	52	5094	Tiotropium/olodaterol	Once daily	Moderate to very severe COPD (post-bronchodilator FEV1/FVC <0.7; and FEV1 <80% pred)	64.0	72.9	37.0	>10	50.0	3
NCT01316887; Donohue [ 58 ]	Multicentre, randomised, double-blind, placebo-controlled, parallel-group	52	177	Umeclidinium/vilanterol	Once daily	COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 ≥35% and ≤80% pred)	61.6	63.9	NA	41.5	54.6	4
NCT01120691; Wedzicha [ 59 ]	Multicentre, randomised, double-blind, parallel-group	64	1108	Glycopyrronium/indacaterol	Once daily	Severe to very severe COPD (post-bronchodilator FEV1/FVC <0.7; FEV1 <50% pred)	63.3	75.0	36.6	>10	37.2	5

Data are presented as n, unless otherwise stated. FEV₁: forced expiratory volume in 1 s; COPD: chronic obstructive pulmonary disease; FVC: forced vital capacity; NA: not available.

Overall, the results obtained from 20 329 COPD patients (LABA/LAMA FDCs, n=9292; LABAs and LAMAs administered as monocomponents, n=11 037) were extracted from 14 studies that reported 20 RCTs. Five studies (nine RCTs) lasted 3 months, six studies (seven RCTs) lasted 6 months and three studies (four RCTs) lasted 12 months.

Pair-wise meta-analysis

LABA/LAMA FDCs significantly (p<0.001) improved trough FEV₁ (95% CI) compared with monocomponents at 3 months (versus LABAs: +109.02 (86.36–131.68) mL, I² 38%, p>0.05; versus LAMAs: +58.05 (38.70–77.40) mL, I² 56%, p<0.01), 6 months (versus LABAs: +72.65 (53.81–91.49) mL, I² 60%, p<0.05; versus LAMAs: +40.53 (15.15–65.91) mL, I² 78%, p<0.001) and 12 months (versus LABAs: +80.51 (66.81–94.20) mL, I² 17%, p>0.05; versus LAMAs: +55.65 (47.26–64.05) mL, I² 0%, p>0.05) of treatment, although LAMAs were significantly (p<0.05) superior to LABAs at 3 and 12 months (figure 1 and supplementary figures S3 and S4).

FIGURE 1.

Forest plot meta-analysis of the impact of long-acting β₂-agonist (LABA)/long-acting muscarinic antagonist (LAMA) fixed-dose combinations on changes in trough forced expiratory volume in 1 s (FEV₁) at 3, 6 and 12 months. Data are presented as mean difference (95% CI) versus monocomponents.

Overall, the LABA/LAMA FDCs were always significantly (p<0.001) more effective in improving SGRQ, when compared with monocomponents (3 months: −2.04 (−2.59–−1.49), I² 0%, p>0.05; 6 months: −1.45 (−1.92–−0.98), I² 8%, p>0.05; 12 months: −0.96 (−1.71–−0.22), I² 62%, p<0.05). In addition, LABA/LAMA FDCs significantly (p<0.001) improved TDI at all time points (3 months: +0.67 (0.45–0.89), I² 0%, p>0.05; 6 months: +0.41 (0.30–0.51), I² 0%, p>0.05; 12 months: +0.39 (0.21–0.56), I² 59%, p<0.05), when compared with monocomponents (figure 2 and supplementary figures S5–S10).

FIGURE 2.

Forest plot meta-analysis of the impact of long-acting β₂-agonist (LABA)/long-acting muscarinic antagonist (LAMA) fixed-dose combinations on a) St George's Respiratory Questionnaire score (SGRQ) and b) transition dyspnoea index (TDI) at 3, 6 and 12 months. Data are presented as mean difference and 95% CI versus monocomponents.

Network meta-analysis

LABA/LAMA FDCs significantly (p<0.001) improved trough FEV₁ (95% CrI) compared with monocomponents at 3 months (versus LABAs: +103.53 (73.40–134.74) mL; versus LAMAs: +62.82 (38.10–87.09) mL), 6 months (versus LABAs: +66.70 (41.45–90.87) mL; versus LAMAs: +38.00 (13.04–62.19) mL) and 12 months (versus LABAs: +81.21 (67.22–95.77) mL; versus LAMAs: +55.78 (45.90–65.91) mL).

The LABA/LAMA FDCs were significantly (p<0.001) more effective in improving SGRQ (95% CrI), when compared with monocomponents, at 3 months (versus LABAs: −1.90 (−3.21–−0.65); versus LAMAs: −1.99 (−2.80–−1.14)), 6 months (versus LABAs: −1.57 (−2.28–−0.72); versus LAMAs: −1.33 (−2.08–−0.59)) and 12 months versus LABAs (−1.84 (−3.63–−0.20)), but not versus LAMAs (−0.62 (−1.99–0.41)).

In addition, LABA/LAMA FDCs significantly (p<0.001) improved TDI (95% CrI) at 3 months (versus LABAs: +0.68 (0.20–1.17); versus LAMAs: +0.64 (0.21–1.09)), 6 months (versus LABAs: +0.42 (0.29–0.57); versus LAMAs: +0.37 (0.23–0.52)) and 12 months (versus LABAs: +0.59 (0.20–1.01); versus LAMAs: +0.33 (0.01–0.67)), when compared with monocomponents.

Detailed results of network meta-analysis, including variance and inconsistency models, are reported in supplementary table S2.

LABA/LAMA FDCs always showed the highest probability of being the best therapy with regard of trough FEV₁, SGRQ and TDI at all time points (overall 98.67%) as confirmed by SUCRA (overall 99.28%), whereas LAMA and LABA as monotherapy were ranked the second and third therapies (table 3).

TABLE 3.

Probability of treatments being the best therapy and surface under the cumulative ranking curve (SUCRA) values

	Probability of being the best therapy	SUCRA value
Trough FEV 1
3 months
LABA/LAMA	100	100
LABA	0	0.5
LAMA	0	49.5
6 months
LABA/LAMA	100	100
LABA	0	1
LAMA	0	49
12 months
LABA/LAMA	100	100
LABA	0	0
LAMA	0	50
SGRQ
3 months
LABA/LAMA	99	99.5
LABA	1	26.5
LAMA	0	24
6 months
LABA/LAMA	100	100
LABA	0	15
LAMA	0	34.5
12 months
LABA/LAMA	91	95
LABA	1	3.5
LAMA	7	50.5
TDI
3 months
LABA/LAMA	100	100
LABA	0	37.5
LAMA	0	12.5
6 months
LABA/LAMA	100	100
LABA	0	14
LAMA	0	36
12 months
LABA/LAMA	98	99
LABA	1	4.5
LAMA	2	47

Data are presented as %. FEV₁: forced expiratory volume in 1 s; LABA: long-acting β₂-agonist; LAMA: long-acting muscarinic antagonist; SGRQ: St George's Respiratory Questionnaire; TDI: transition dyspnoea index.

Heterogeneity, publication bias and optimal information size

Although substantial heterogeneity due to the small-study effect was detected for some summary effects, Egger's test did not indicate any significant (p>0.1) publication bias (supplementary figure S11). In effect, in contrast with the overall test of heterogeneity, Egger's test represents a powerful tool for assessing a specific type of heterogeneity leading to publication bias. However, since any analysis of heterogeneity depends on the number of RCTs included in the analysis, and meta-analysis with small number of studies may limit the statistical power of the test, we based evidence for bias on p<0.1.

In addition, our meta-analysis met a reasonable optimal information size to ensure a very good (probability of observing 30% overestimation for τ² =0.25: <1% at true relative risk reduction 10%) to excellent (probability of observing 20% overestimation for τ² =0.05: <1% at true relative risk reduction 0%) low risk of observing an overestimated intervention effect due to random errors, and the exact numbers of optimal information size for all outcomes and at each time point are provided in supplementary table S3.

Sensitivity analysis

The sensitivity analysis indicated that the results obtained in our network meta-analysis were consistent with those of our pair-wise meta-analysis (GMD: overall 4.25%, FEV₁ 5.65%, SGRQ 4.34% and TDI 2.76%).

In contrast, significant discrepancy (p<0.05) was detected between the results of our meta-analysis and that of Oba et al. [4]. The overall GMD was 26.68% (FEV₁ 19.02%, SGRQ 19.39% and TDI 41.62%), and the overall discrepancy increased to 43.33% (FEV₁ 32.91%, SGRQ 32.60% and TDI 64.49%) when the data missed by Oba et al. [4] at 12 months were included into the analysis.

Discussion

The results of our meta-analysis, which has not been affected by any publication bias, support the evidence that the dual bronchodilation with a LABA and a LAMA is more effective than monocomponents [5–8] after 3, 6 and 12 months of treatment. In fact, the improvement of trough FEV₁ induced by LABA/LAMA FDCs versus monotherapy was greater than the minimal clinically important difference (MCID) of 60 mL between active comparators (as defined by the European Medicines Agency guidelines on clinical investigation of medicinal products in the treatment of COPD), at least with regard to LABAs administered as monocomponents [60]. Although some of the analysed RCTs investigated the improvement of SGRQ scores and TDI induced by LABA/LAMA FDCs versus monotherapy, to date no published consensus exists for the MCID thresholds between active comparators for these variables. Thus, it was expected that the effect estimates resulting from our meta-analysis indicated that LABA/LAMA FDCs were unable to improve SGRQ scores and TDI more than the MCIDs (four units and one unit, respectively), when compared with monocomponents. In any case, we cannot rule out that specific LABA/LAMA FDCs may have different influences on the proportion of responder patients with regard of SGRQ and TDI, when compared with monotherapy.

Indeed, this study represents the natural step forward from our previous meta-analysis [5], in which we identified a gradient of effectiveness among the currently approved LABA/LAMA FDCs. The novel findings provided here support evidence that the dual bronchodilation by LABA/LABA FDCs is more effective than that elicited by monocomponents, especially with regard to LABAs for changes from baseline in trough FEV₁. Furthermore, the overall superiority of LABA/LAMA FDCs was greater after 3 months of treatment, while it diminished after 6 and 12 months of treatment. This trend suggests that continued improvements in FEV₁ elicited by LABA/LAMA combinations can be expected over the first 3 months of treatment; after that, the greater benefits of dual bronchodilation remains stable. Thus, it seems that for long-acting bronchodilator agents the time taken to reach the clinical bronchorelaxant steady state is considerably longer than the time taken to achieve the pharmacodynamic steady state [61, 62], meaning that the LABA/LAMA interaction is fundamental, not only after acute administration [7, 8], but also over time in the course of chronic treatment.

We must highlight that our results are considerably different from those obtained by Oba et al. [4], with an overall discrepancy of ≈26% with regard to FEV₁, SGRQ and TDI variables. This incongruity reached ≈43% if the data missed by Oba et al. [4] at 12 months were included in the analysis. Although we cannot understand why data at 12 months were omitted, we suppose that the relevant inconsistencies between our results and those published by Oba et al. [4] may be related to an inadequate selection of RCTs, an inappropriate extraction of raw data from published papers and an inaccurate consultation of repository databases such as clinicaltrials.gov. In addition, the authors [4] used a sophisticated network meta-analytic approach that may have led to frequent inconsistent and biased results [63].

However, in order to dispel any doubt, we have performed a sensitivity test by performing a network meta-analysis on the 14 studies included in the quantitative synthesis. Intriguingly, we have detected only ≈4% difference when compared with our pair-wise meta-analysis. This finding suggests that the distance between our results and those of Oba et al. [4] cannot be strictly attributable to the difference between the pair-wise and network meta-analytic approach.

We fully agree with Deeks et al. [64] that the production of a diamond at the bottom of a plot is an exciting moment for any researchers. Nevertheless, results of meta-analyses can be very misleading if adequate attention has not been given to formulating the review question, specifying eligibility criteria, identifying, selecting and critically appraising studies, collecting appropriate data and deciding what would be meaningful to analyse [64].

In the light of these evidences, and considering the inaccurate network meta-analysis performed by Oba et al. [4], our pair-wise meta-analysis provides an alternative view of the examination of dual bronchodilation with regard to the duration of treatment, based on the evaluation of the pharmacological characteristics of bronchodilator agents and performed through simpler meta-analytic methods.

We strongly believe that the effectiveness of dual bronchodilation should always be compared with at least one of the monocomponents included in the FDCs, and not with different LABAs or LAMAs. In fact, it should be obvious that comparing LABA/LAMA combinations with monocomponents characterised by different pharmacokinetics and pharmacodynamics represents a strict pharmacological matter [65, 66], and not a merely statistical problem that may be solved by performing a subgroup analysis. Furthermore, we believe that the regimen of administration should be consistent between the drugs of each combination, i.e. both medications administered once daily or twice daily, and not at different time intervals.

Finally, several LABA/LAMA FDCs have been approved for COPD, but no head-to-head comparative RCTs have yet been conducted. We strongly believe that the real gradient of effectiveness of the approved LABA/LAMA FDCs remains an unmet medical need, which can be indirectly investigated using specific and well-performed meta-analytic approaches, considering that a homogenous body of literature already exists.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary data ERR-0043-2016_Supplementary_data ^{(1.5MB, pdf)}

Disclosures

L. Calzetta ERR-0043-2016_Calzetta ^{(1.2MB, pdf)}

M. Cazzola ERR-0043-2016_Cazzola ^{(1.2MB, pdf)}

M.G. Matera ERR-0043-2016_Matera ^{(1.2MB, pdf)}

E. Puxeddu ERR-0043-2016_Puxeddu ^{(1.2MB, pdf)}

P. Rogliani ERR-0043-2016_Rogliani ^{(1.2MB, pdf)}