This comprehensive and up-to-date evaluation of the effects of beta-blockers in patients with COPD adds to the previous literature in several ways: we included all studies reporting on any type of beta-blocker treatment in patients with COPD, showing overall beneficial effects on AECOPD and mortality. For the first time, we used a probabilistic approach to evaluate the effect of beta-blockers on FEV1 using direct and indirect evidence from RCTs in an NMA, comparing seven treatments against placebo, and presented results for patients with COPD with and without CVD disease separately. No beta-blocker affected lung function significantly except propranolol, and the treatments less likely to have a detrimental effect on FEV1 were labetalol (in those with COPD and CVD) and celiprolol (in those with COPD without explicit CVD). Lastly, we found that data on all-cause hospitalization and quality of life endpoints such as SGRQ, 12 and 6MWT and SF-36 were scarcely reported across the literature and did not lend themselves to formal quantitative analysis—suggesting an area of focus for future studies.
Mortality
Despite heterogeneous elements such as follow-up time, baseline characteristics including age, sex and comorbidities and geographical location, individual results from the 17 out of 21 studies reporting on mortality suggested beta-blocker therapy was associated with a diminished risk of death compared to those not prescribed beta-blockers, in patients with COPD. However, this quality of evidence was deemed “low” per GRADE assessment (Additional file
1: Table S17) and we were not able to quantify the effect of beta-blockers on mortality due to considerable heterogeneity (I
2 > 75%). Previous reports [
10,
11,
67] have provided pooled estimates of reductions in mortality risk associated with beta-blocker treatment, however all reported degrees of heterogeneity above the Cochrane I
2 threshold of 75%; 89.3% [
10], 83% [
11] and most recently 96% [
67] bringing into question the validity and interpretability of these results as applied to the general COPD population. Reasons for very high heterogeneity in previous meta-analyses include: differences in study populations (i.e. including patients with differing degrees of severity), inaccurate risk of bias assessment and inclusion of different comparators for the intervention effect of interest (i.e. including studies where comparator arms received calcium channel blockers, despite aiming to assess the effect of beta-blocker treatment vs. lack of treatment) [
67].
In our analysis, most studies were affected by bias, particularly due to confounding: two studies did not adjust for any covariate factors [
27,
55], whilst nine did not adjust for COPD severity either directly, or indirectly by including COPD medication regimen/exacerbation history in the final model [
25‐
28,
30,
32,
34,
36,
37]. Therefore, these studies may overestimate the prognostic effect of beta-blocker therapy on patients with COPD and may, in turn, skew results to show benefits. One of the reasons for the lack of adjustment for COPD-related variables may be due to using data from either existing drug-trials or CVD-specific registries which included data on subgroups of patients with COPD, reiterating the need for trials designed specifically for patients with COPD (with and without additional CVD) which may allow for reliable assessment of the true effect of beta-blockers in these patients. Furthermore, it is not surprising to observe a decrease in mortality, as this could be related to the effect of beta-blockers on other comorbid conditions of patients (i.e. CVD), which is established. A previous study [
33] suggested long-term treatment with beta-blockers improved survival of patients with COPD without CVD, however future studies are needed to confirm this result and to assess whether beta-blockers provide non-CV mortality benefits.
AECOPD
We found evidence to suggest that patients with COPD who are given beta-blockers are at decreased risk of AECOPD (HR 0.78 [95%CI 0.74–0.82]), replicating findings from Du and colleagues [
10] who report an even larger reduction in risk, of 37% (RR 0.63 [95% CI, 0.57–0.71]). However, this previous meta-analysis, had methodological limitations inherent to the observational nature of the pooled studies (i.e. residual confounding, immortal time bias), which may limit generalizability of results. However, the GRADE assessment revealed the body of observational evidence on which our estimate was derived was of “low” quality (Additional file
1: Table S19). A recent RCT [
12], less likely to be affected by the biases of previous observational studies, found no significant difference between metoprolol and placebo on the time to AECOPD of any severity, but revealed a significant increase in risk of AECOPD requiring hospitalization, in patients with COPD without an indication for beta-blocker treatment, bringing into question the protective effect of this specific beta-blocker agent.
However, this trial did not evaluate other beta-blockers, therefore future RCTs evaluating multiple regimens, are needed to confirm the benefit of these agents. Whether beta-blockers have an indirect effect on exacerbations of COPD could be assessed in clinical trials including patients with COPD and comorbid CVD, allowing assessment of these agents in a more representative COPD population.
FEV1
FEV1 was assessed in 199 patients enrolled in 12 RCTs and we found that none of the individual cardioselective beta-blockers included in our NMA (atenolol, bisoprolol, celiprolol, metoprolol) were associated with significant effects on lung function in patients with COPD, regardless of baseline FEV1 or follow-up time. This is in line with a Cochrane review [
9] which concluded that cardioselective beta-blockers given in either single dose or for longer durations, do not affect FEV1 in patients with COPD, even in those with the lowest baseline FEV1 measurements. Furthermore, our report extends to incorporate a lack of effect on FEV1 of non-selective beta-blockers such as carvedilol and labetalol. Propranolol was the only medication found to be associated with a reduction of 140 ml in FEV1 (95% CrI: -0.28, -0.016), which is larger than the threshold of 100 ml change deemed clinically significant by the American Thoracic Society and European Respiratory Society guidelines. This result is based on high quality evidence, according to the GRADE assessment (Additional file
1: Table S19), and thus supports current recommendations to not use this medication in patients with COPD.
For the first time reported in the literature, we aimed to rank beta-blockers with respect to their effect on lung function. Propranolol had the lowest probability of being ranked first (suggesting worse impact on lung function), compared to all other individual treatments considered in our NMA, including placebo. Labetalol and celiprolol—drugs used in hypertension—were the least likely drugs to negatively impact FEV1, compared to all other beta-blockers; however, neither affected FEV1 with certainty compared to placebo and results were inferred from very low quality evidence according to GRADE (Additional file
1: Table S18), bringing into question their leading positions in the hierarchy. Since choice of beta-blocker may be influenced by CVD comorbidity (i.e. carvedilol, metoprolol and bisoprolol are recommended in stable HF; atenolol is more often prescribed in patients with asymptomatic hypertension, while bisoprolol is also used in atrial fibrillation, and propranolol is infrequently used to treat tachyarrhythmias), it is perhaps not surprising that we did not identify a clear “best” beta-blocker to be used in COPD. The fact that the beta-blockers less likely to decrease lung function are mainly used to treat hypertension may just reflect this subgroup of patients could be less prone to detrimental side-effects (i.e. indication bias), compared to others with COPD and more severe comorbidities. Indeed, the prescription of beta-blockers in COPD needs to consider clinically significant lung function alteration vs. mortality benefits in those with CVD, particularly MI [
68] and HF [
69].
Whilst CVD is diagnosed in 20 to 60% patients with COPD [
70], our main analysis included primarily small trials and only three explicitly included patients with a cardiac comorbidity (one included angina [
54], two included HF patients [
55,
57], and one included patients with hypertension, which is a common CVD risk factor [
50]. In line with previous research [
9], we report no significant FEV1 treatment effect in patients with COPD with CVD.
The remaining eight trials excluded those with CVD (or simply did not report whether this was present), and results mirrored those observed for patients with CVD. Whilst results from this subgroup analysis are encouraging, previous clinical data on in this subgroup is scarce. A recent single RCT including COPD patients without an indication for beta-blockers (therefore those with HF, previous MI or revascularization) failed to demonstrate clear benefits of metoprolol over placebo. Observational studies have included a more varied breadth of specific beta-blockers, however they do not present a clear picture: the population-based Rotterdam Study [
71] reported significant decreases in FEV1 associated with both cardio and non-cardioselective beta-blockers, while two other studies, one from Scotland [
35] and an one from Japan [
72] reported no significant difference in FEV1. Yet, these results may be affected by confounding by indication, which may explain the variability of estimates. Additionally, the longer follow-up times in these studies (ranging from 4 to 6 years) may overlook effects of FEV1 decline which is documented in patients with COPD, regardless of CVD comorbidities.
Overall, our FEV1 analysis suggests the beta-blockers included in this review do not affect lung function in patients with COPD regardless of CVD disease status, and selectivity of agent does not appear to have an impact. However, the two treatment networks contained different medications (celiprolol was assessed in one trial excluding CVD, while labetalol in one trial including CVD) thus we cannot rule out any other potential differential results if a whole range of beta-blockers were included. Finally, we included evidence based on a relatively small population and some of the studies were conducted decades ago; therefore, large clinical studies are needed to assess other agents which may confer lung function benefits across contemporary COPD patients.
The effect of beta-blocker exposure on all-cause hospitalization and quality of life outcomes in patients with COPD could not be quantified, due to a paucity of data. Narrative results from the assessment of studies investigating quality of life outcomes, such as SGRQ, 12 and 6 MWT and SF-36 all suggest non-significant effect of beta-blockers, from both RCTs and observational studies, albeit the data was deemed to be of “very low” quality according to GRADE (Additional file
1: Table S17). Currently, COPD management is focused on preventing exacerbations and improving functioning and health-related quality of life. Clinical studies of beta-blocker treatment in cardiac disease suggests improvements in exercise tolerance and functional status, so whether beta-blockers impair or improve these outcomes in patients with COPD also, is a topic of importance for clinical management. Both randomized trials and, importantly, prospective observational studies with longer follow-up times are needed.
Limitations
There are several limitations to our analysis: first, we included published, peer-reviewed literature only thus, results may affected by publication bias as it is more likely that studies reporting positive results (i.e. that did not find beta-blockers were associated with negative outcomes) are more often reported than negative studies. Nevertheless, our data is based on the most recent available evidence and portray a nuanced implication of specific beta-blocker treatment in patients with COPD, emphasizing the need for a targeted treatment of CVD comorbidity in these patients.
We only included stable COPD patients and whilst we showed that FEV1 reduction (or increase) was not significant according to beta-blocker exposure (apart for propranolol), we could not verify whether these therapeutic agents diminish the response to rescue COPD medication such as beta-agonists, administered during an exacerbation of COPD. We also did not verify long-term effects of co-administration of beta-blockers and beta-agonists and how their interaction may affect outcomes in patients receiving both types of medication.
Another issue is undiagnosed CVD in patients with COPD. Symptoms of ischemic heart disease or HF may be misattributed or overlapping with COPD, and thus not formally diagnosed, posing difficulties in disentangling possible non-cardiac effects of beta-blockers, independent of their proven cardiac benefits. One advantage of our FEV1 analysis is that we included RCTs only, where concomitant CVD is often ascertained more rigorously and therefore CVD status was known with a greater degree of confidence that may be the case in observational studies.
Furthermore, no statistically significant effect was detected in subgroup analyses stratified by CVD status, which may be due to limited sample size. Future, adequately powered RCTs are needed to assess the effect of beta-blockers in a diverse COPD population, allowing for accurate comparisons based on CVD status to be made.
A recent RCT [
12] comparing metoprolol with placebo failed to find a significant effect on FEV1, but reported worsening of dyspnea and overall COPD symptoms, suggestive of respiratory effects not captured by spirometry. This confirms the need to evaluate a spectrum of respiratory outcomes to fully assess the implications of beta-blocker treatment in patients with COPD, which needs to be addressed in future studies
Confounding by contraindication is likely to affect interpretation of results—if we assume clinicians knowingly withheld treatment from patients due to concerns regarding breathlessness, this may have resulted in a reduced sample size of possible COPD patients who may have been eligible for beta-blocker therapy. Alternatively, doctors may prescribe beta-blockers to less severe patients, limiting generalizability.
Our AECOPD analysis is also limited by a low number of included studies, all of which were observational—we identified one RCT only (evaluating metoprolol). This reinstates the need of more carefully conducted RCTs to evaluate a range of beta-blockers and their effects of AECOPD, in order to validate observational data.