Abstract
Introduction Inhaled antibiotics may improve symptom scores, but it is not known which specific symptoms improve with therapy. Item-level analysis of questionnaire data may allow us to identify which specific symptoms respond best to treatment.
Methods Post hoc analysis of the AIR-BX1 studies and two trials of inhaled aztreonam versus placebo in bronchiectasis. Individual items from the quality of life bronchiectasis (QOL-B) respiratory symptom scale, were extracted as representing severity of nine distinct symptoms. Generalised linear models were used to evaluate changes in symptoms with treatment versus placebo from baseline to end of first on-treatment cycle and mixed models were used to evaluate changes across the full 16-week trial.
Results Aztreonam improved cough (difference 0.22, 95% CI 0.08–0.37; p=0.002), sputum production (0.30, 95% CI 0.15–0.44; p<0.0001) and sputum colour (0.29, 95% CI 0.15–0.43; p<0.0001) versus placebo equating to a 20% improvement in cough and 25% improvement in sputum production and colour. Similar results were observed for cough, sputum production and sputum purulence across the trial duration (all p<0.05). Patients with higher sputum production and sputum colour scores had a greater response on the overall QOL-B (difference 4.82, 95% CI 1.12–8.53; p=0.011 for sputum production and 5.02, 95% CI 1.19–8.86; p=0.01 for sputum colour). In contrast, treating patients who had lower levels of bronchitic symptoms resulted in shorter time to next exacerbation (hazard ratio 1.83, 95% CI 1.02–3.28; p=0.042).
Conclusion Baseline bronchitic symptoms predict response to inhaled aztreonam in bronchiectasis. More sensitive tools to measure bronchitic symptoms may be useful to better identify inhaled antibiotic responders and to evaluate patient response to treatment.
Abstract
Inhaled aztreonam improves cough and sputum production in patients with bronchiectasis but has no impact on other symptoms. More sensitive tools to measure bronchitic symptoms may be useful to enrich for responders and to evaluate patient benefit. https://bit.ly/2UMKM5i
Introduction
Bronchiectasis significantly impacts upon patients’ quality of life, affecting physical, emotional and social aspects of wellbeing [1, 2]. Disease progression is driven by recurrent exacerbations, which are associated with reduced lung function and in severe cases can lead to respiratory failure and even death [1, 3–5]. Exacerbation frequency and long-term mortality are increased in patients with chronic airway infection with bacterial pathogens [6]. Pseudomonas aeruginosa is the most common organism isolated from patient sputum worldwide and is strongly associated with worse quality of life, lower lung function, higher risk of hospitalisation and increased mortality rates [7, 8]. This high burden of disease highlights the need to develop evidence-based therapies that can reduce the burden of bacterial infection. Inhaled antibiotics deliver high concentrations of antibiotic to the site of infection resulting in marked reductions in airway bacterial load in patients with bronchiectasis [9]. However, it has been challenging to demonstrate clearly that these reductions in bacterial burden translate into clinical benefits. A recent meta-analysis of 16 trials including 2597 patients with bronchiectasis showed strong antimicrobial efficacy with a pooled reduction of 2.3 log units in colony-forming units, but this translated into only modest reductions in exacerbation frequency (rate ratio 0.8, 95% CI 0.67–0.97) and no clinically significant improvement in symptoms [9]. Despite these inconclusive data, inhaled antibiotics are still frequently used “off-label” with patient and clinician perception that they provide benefits [10].
A possible explanation for this difference between “real life” and randomised studies is the tools used to measure symptoms [11]. The number of validated health-related quality of life tools used in bronchiectasis trials is limited, and the quality of life bronchiectasis questionnaire (QOL-B) is the only disease-specific tool that has been tested in multiple randomised trials [11]. Its respiratory symptom scale consists of nine questions asking about congestion, cough, sputum production, sputum colour, shortness of breath, wheezing, chest pain, shortness of breath while talking and nocturnal cough. Each of these symptoms is potentially important to patients with bronchiectasis, and each symptom is likely to be associated with severity of disease and morbidity [2, 11]. However, it is less certain that a specific treatment would be expected to improve all of these symptoms.
Extensive prior research suggests that elevated bacterial load induces neutrophil inflammation, including the release of myeloperoxidase which causes sputum purulence and neutrophil elastase which is a key mediator of cough and sputum production through direct stimulation of mucin production from epithelial cells and impairment of mucociliary clearance [12–14]. Inhaled antibiotic treatment has been shown to rapidly reduce neutrophilic inflammation and therefore might be expected to also reduce cough, sputum production and sputum purulence [15]. In contrast, breathlessness is associated with airflow obstruction, emphysema, cardiovascular fitness, anaemia conditioning and neuromuscular function, all of which may not be immediately modified with inhaled antibiotics [3, 16]. In the recent meta-analysis, inhaled antibiotics reduced, rather than improved forced expiratory volume in 1 s (FEV1) [9].
We hypothesised that by examining the effect of an inhaled antibiotic on individual symptom responses we would observe that inhaled antibiotics improve cough, sputum and sputum purulence while having much less effect on other symptoms. To investigate this, we used itemised QOL-B data collected during two large, previously published, randomised controlled trials of inhaled aztreonam lysine (AZLI) [17].
Methods
AIR-BX1 and AIR-BX2 were identical, double-blind, multicentre, randomised, placebo-controlled, phase 3 trials consisting of two 4-week-on, 4-week-off treatment cycles. Patients received either inhaled AZLI 75 mg or placebo three times daily. Eligibility criteria has been published previously [17]. This analysis only included patients who gave additional consent for future exploratory analysis of their data and we have previously reported that there were no significant differences in characteristics between the original trial cohort and those who consented to future analysis [18].
The primary end-point used in the AIR-BX studies was change in QOL-B respiratory symptoms score (RSS) at week 4, the end of the first on-treatment cycle [11]. Secondary end-points included the change in QOL-B-RSS at week 12. The minimal important difference for the QOL-B-RSS as a whole is an 8-point change [11].
Time to first protocol-defined exacerbation by week 16 was a secondary end-point of special interest in the AIR-BX studies. The exacerbation definition used was as follows: acute worsening of respiratory disease meeting at least three major criteria (or two major and at least two minor). Major criteria were increased sputum production, change in sputum colour, dyspnoea and cough, while minor criteria were fever (>38°C) at clinic visit, increased malaise or fatigue, FEV1 (L) or forced vital capacity decreased by >10% from baseline, and new or increased haemoptysis [17].
Item-level analysis
The nine items of the QOL-B-RSS are shown in table 1. The items are answered with a 1-week recall period and have four options. Lower scores reflect poor quality of life from increased respiratory symptoms, while high scores show lesser impact. For individual item-level analysis, the responses (table 1) were converted into numerical values for analysis, with 1 representing the most severe response (a lot/always) and 4 representing the least severe response (not at all/never). For sputum colour, there are six possible responses in the QOL-B. “Brownish-dark” and “green with traces of blood” are both rated the same severity and no score is given to the answer “Don't know”.
Individual items completing the quality of life bronchiectasis (QOL-B) respiratory symptom domain. These questions form the last domain of QOL-B and are numbers 29–37
Statistical analysis
Changes from baseline to visit 4 (end of first on-treatment period)
To test the hypothesis that AZLI treatment would improve specific symptoms linked to bacterial load we analysed the effect of AZLI treatment versus placebo on each individual item within the QOL-B RSS questionnaire as listed in table 1. Analysis was conducted using the intention-to-treat principle. Studies were initially evaluated separately (AIR-BX1 and AIR-BX2) and then pooled. The second hypothesis was that patients with more cough, sputum production and sputum purulence at baseline would experience a superior response to inhaled antibiotic treatment. In the analysis of change from baseline to visit 4 we used a generalised linear model with the change in each individual symptom domain as the outcome in each of the four baseline severity categories. High-symptom patients were those answering 1–2 per item and low-symptom patients were those answering 3–4.
Repeated measures analysis across the entire study duration
For analysis of whether baseline symptoms predicted QOL-B-RSS response across the entire trial duration we used a mixed effects model. The full model for each question contained the four treatment terms: 1) an immediate effect of treatment (visits 3, 4 and 6 while patients were receiving either AZLI or placebo) and 2) a longer term (study duration incorporating data from visits when off-drug) placebo effect affecting both arms equally, as well as 3) an additional short-term and 4) a longer term drug effect that only affected the patients receiving AZLI. For each analysis data are presented for full models, which contain all four treatment parameters, and for a best model in which all available parameters were statistically significant. The data from the two trials was combined, and an intercept plus a single common drift term included along with individual as a random effect. Interactions were included for baseline symptoms and AZLI response. Model selection and averaging was done by Akaike information criterion.
To evaluate treatment effects of individual symptoms across the entire study duration we used a similar mixed effects model with the response to each individual question as the outcome. As there are only four possible answers to each question, uncertainties around these models' estimates were generated by bootstrap resampling the data. The data from the two trials was pooled for simplicity and because preliminary analyses identified no differences between the response in the two trials. 15 submodels were fitted, containing each combination of the four treatment terms. Terms were considered statistically significant where their 95% confidence interval from bootstrapping did not cross zero. Among those for which all parameters were statistically significant, the model for each question that contained the largest number of parameters was considered best.
For interpretation of results we considered a p-value of <0.05 as statistically significant. In view of the multiple comparisons performed we also applied a Holm–Bonferroni correction to the primary results. All statistical analysis was performed using IBM SPSS Statistics (version 25; IBM, Armonk, NY, USA) or R version 3.5.1 (www.r-project.org).
Results
AIR-BX1 and AIR-BX2 demographics
A combined 440 participants were included in the analysis. Table 2 shows the demographics for participants. More detailed patient characteristics have been published previously, demonstrating no significant differences between the patient characteristics in AIR-BX1 and AIR-BX2 [18]. Patients had an average age of 64 years. Patients were predominantly female (70%) with moderate FEV1 impairment. 131 protocol-defined exacerbations were recorded throughout the study, although 70% of participants did not experience an exacerbation during the 3-month study.
Combined AIR-BX1 and AIR-BX2 participant characteristics
Treatment response for individual symptoms
First we examined the original trial primary end-point, which was change from baseline to visit 4, after 28 days of AZLI treatment or placebo. For the total score, the original trials reported an improvement of 0.8 points (95% CI −3.1–4.7, p=0.68) in AIR-BX1 and 4.6 points (95% CI 1.1–8.2, p=0.011) in AIR-BX2. For individual symptoms we observed clear improvements at week 4 in daily cough, sputum production and sputum purulence of 20%, 25.7% and 25.2%, respectively (all p<0.05; table 3). In addition, there were nonsignificant improvements in congestion, but negative effect estimates for breathlessness, wheeze and chest pain which were not statistically significant. Supplementary table S1 shows the results for each individual study. Consistent with the original trials, the effects were stronger in AIR-BX2, but were concordant in the sense that more patients achieved a ≥2-point improvement in individual symptoms with AZLI than with placebo by week 4 for congestion (p=0.015), sputum production (p=0.003) and sputum purulence (p=0.01). No significant differences were observed between groups for the other symptoms (supplementary table S2).
Pooled item-level response following 4 weeks of treatment with aztreonam lysine (AZLI) versus placebo
The mixed model repeated measures analysis supported the results of the week-4 analysis. Figure 1 shows the trajectories of individual symptoms from baseline to the end of trial. We observed improvements in symptoms for both AZLI- and placebo-treated subjects consistent with a placebo effect. A clear treatment benefit was evident in figure 1b–d consistent with improved cough, sputum production and sputum colour. Interestingly, inhaled antibiotic treatment appeared to provide a sustained improvement in cough and sputum production during the off-treatment period, but sputum colour returned to baseline during the off-treatment period.
Trajectories of individual symptoms comparing aztreonam lysine (AZLI) treatment versus placebo in the full model as described in the statistical analysis section. Seven visits are shown; visits 1 and 2 are screening and baseline prior to treatment, visit 3 is 14 days into the first treatment cycle, visit 4 is at 28 days (end of the first treatment cycle), visit 5 is at the end of the off-treatment cycle (28-days off treatment). Visit 6 is the end of the second 28-day treatment cycle, while visit 7 is the end of the final 28-day off-treatment cycle and is the end of the study. a) Question 29, congestion; b) question 30, cough; c) question 31, sputum production; d) question 32, sputum colour; e) question 33, shortness of breath; f) question 34, wheeze; g) question 35, chest pain; h) question 36, shortness of breath when talking; i) question 37, nocturnal cough. QOL-B-RSS: quality of life bronchiectasis respiratory symptom scale.
Results of the mixed models are shown in supplementary table S3. The best mixed model found a statistically significant improvement in cough and sputum with AZLI treatment throughout the study (difference 0.11, 95% CI 0.007–0.23 for cough and 0.19, 95% CI 0.09–0.29 for sputum production), and a significant improvement in sputum colour on AZLI treatment at visits 3, 4 and 6 (difference 0.25, 95% CI 0.17 to 0.33). In contrast, while on AZLI treatment there was a significant deterioration in breathlessness (−0.09, 95% CI −0.17−0.002) and wheeze (−0.11, 95% CI −0.20−0.02) (supplementary table S3).
We conclude that examining both the primary end-point of week 4 and the trajectory across the study, inhaled aztreonam treatment improved cough, sputum production and sputum colour, but resulted in worsening of breathlessness and wheeze.
Do baseline symptoms predict treatment response?
QOL-B RSS
The improvement in QOL-B RSS from baseline to 4 weeks observed in the overall population was associated with baseline symptoms. Patients with more severe sputum production and sputum colour had a greater response in terms of QOL-B RSS. For sputum production, those reporting “a lot” and a “moderate amount” (n=309, 70.2%) had a statistically significant improvement (difference 4.82, 95% CI 1.12–8.53; p=0.011), while those reporting “a little” or “not at all” had no treatment benefit (difference −2.61, 95% CI −7.58–2.37; p=0.30). Likewise, patients with green or brownish-dark sputum (n=240, 54.5%) had an improvement in QOL-B RSS overall (difference 5.02, 95% CI 1.19–8.86; p=0.01), while those with yellow or clear sputum did not (difference −0.78, 95% CI −5.88–4.33; p=0.77). Results can be seen in supplementary table S4.
Intriguingly, patients reporting little to no wheezing (n=328, 74.5%; difference 3.74, 95% CI 0.50–6.97; p=0.024), shortness of breath when talking (n=338, 76.8%; difference 3.23, 95% CI 0.12–6.33; p=0.042) and nocturnal cough (n=319, 72.5%; difference 3.62, 95% CI 0.27–6.97; p=0.034) at baseline also showed statistically significant beneficial effects of treatment on overall QOL-B RSS. No other significant effects were observed.
In the mixed models across all study visits the only statistically significant interaction was for sputum production, whereby higher sputum production at baseline was significantly associated with QOL-B RSS treatment response across the entire study duration difference 2.18 (95% CI 0.14–4.22, p=0.04) (supplementary tables S5 and S6).
Exacerbations
In the original AIR-BX analysis, AZLI was associated with a nonsignificant shortening of time to first exacerbation and a higher number of exacerbations in the AZLI arm. This was speculated to be due to adverse effects of inhalation and bronchospasm; however, our analysis showed a phenotypical relationship, with patients having a lower severity of congestion (n=207, 47.0%; hazard ratio (HR) 1.69, 95% CI 1.02–2.8; p=0.042), cough (n=152, 34.5%; HR 1.65, 95% 0.95–2.90; p=0.078), sputum colour (n=200, 45.5%; HR 1.83, 95% CI 1.02–3.28; p=0.042) and nocturnal cough (n=121, 27.5%; HR 1.57, 95% CI 1.03–2.37; p=0.034) having a shorter time to first exacerbation. In contrast, although not statistically significant, patients with greater baseline wheeze (n=112, 25.5%) had a shorter time to first exacerbation (HR 1.75, 95% CI 0.92–3.33; p=0.086) (table 4).
The effect of baseline symptoms on time to first exacerbation
FEV1
No effect of treatment was observed on FEV1 when stratified by baseline symptoms (supplementary table S7).
Discussion
The aim of this post hoc analysis was to identify which QOLB symptoms respond to treatment with inhaled aztreonam. We show clear differences in treatment response between different symptoms that make up the respiratory symptom scale of the QOL-B questionnaire. Treatment with AZLI resulted in clear improvements in cough, sputum production and sputum purulence while causing no effect on other symptoms except for shortness of breath and wheeze, which were slightly worse with treatment. Given that these symptoms improve with therapy, it would be expected that patients with more severe cough, sputum production and sputum purulence at baseline would respond better to inhaled antibiotics and this was what we observed. Interestingly, in the AIR-BX studies there was an increase in exacerbations in the treatment group which was driven by an increase in adverse events [17]. We show that patients treated with inhaled antibiotics that lack the above symptoms were more likely to experience exacerbations at an earlier time point.
A series of randomised clinical trials of inhaled antibiotics in bronchiectasis have failed to demonstrate a significant effect on their primary end-points. This leads to a clear need to identify which patients may respond to inhaled antibiotics [19–22]. An understanding of the biology of how inhaled antibiotics work suggests that they should modify some symptoms and not others and that our current methods of identifying treatment response may not be optimal.
Taken together, this study adds to our understanding of the role of inhaled antibiotics in bronchiectasis and, while the post hoc nature of our study requires some caution in the interpretation of their details, the results are concordant with the reported pathophysiology of bronchiectasis. Bacteria, particularly P. aeruginosa produce an intense neutrophil-mediated inflammatory response that increases in proportion to airway bacterial load [23, 24]. Patients with higher bacterial load and higher levels of neutrophilic inflammation measured using cell counts or markers of neutrophil activation such as neutrophil elastase have worse symptoms and a higher frequency of exacerbations [25, 26]. Neutrophil elastase in particular is directly linked to symptoms by provoking secretion of mucins from bronchial epithelial cells, particularly MUC5AC, which is a key mucin in bronchiectasis airway secretions and is linked to disease severity [27, 28]. In addition, neutrophil elastase has been reported to impair mucociliary clearance through direct effects on ciliated epithelium [14]. Neutrophil elastase is released from neutrophil primary granules along with myeloperoxidase, the concentration of which greatly determines the green colour of purulent sputum [23, 25]. Our previous work showed that neutrophil markers reduce in parallel with reducing bacterial load [15, 29]. Therefore, our results verify this model whereby inhaled antibiotics reduce bacterial load, which reduces neutrophilic inflammation which therefore reduces the stimulus for mucin secretion, improves mucociliary clearance, reducing cough and sputum production, and reduces sputum purulence through a fall in myeloperoxidase concentration.
We have recently shown that patients with higher bacterial load at baseline are more responsive to inhaled antibiotics [18]. Currently, bacterial load quantification is not routinely tested in most healthcare environments, whereas identifying clusters of symptoms that predict response is simple and easily implemented into clinical practice. It is perhaps not surprising that if inhaled antibiotics reduce cough, sputum production and sputum purulence that patients with more of these symptoms respond better to treatment. Nevertheless, this is useful information for future design of trials. Despite the QOLB-RSS being used as the primary end-point for the AIR-BX trials, patients were not required to have any specific baseline symptoms for enrolment. When we analysed the baseline data we found that as many as 64.8% of participants did not have the reported symptom at baseline (known as a ceiling effect), and therefore no measurable improvement is possible in that symptom. Despite the study enrolling patients with a history of Gram-negative infection, many patients were not symptomatic and therefore may not have been typical of the kind of patients being prescribed inhaled antibiotics in clinical practice [17].
It is interesting that wheeze and shortness of breath somewhat worsened during treatment. This is consistent with the recent meta-analysis performed by Laska et al. [9], who showed a mean reduction in FEV1 with inhaled antibiotic treatment across 16 trials, and the results of individual studies which have shown that inhaled antibiotics are irritant to the epithelium and therefore can cause bronchoconstriction or bronchospasm. While there has been a lot of focus on the number of patients withdrawing from treatment as a result of bronchospasm, our results suggest that even in patients who persist with therapy there can be a deterioration in shortness of breath.
How do these results affect our future approach to inhaled antibiotic trials? Approval of drugs by regulatory agencies such as the United States Food and Drug administration requires demonstration that drugs improve how patients “feel, function or survive”. This means showing an effect on symptoms or quality of life. We suggest that future studies should aim to enrich for patients with a higher level of the symptoms that are likely to respond to inhaled antibiotics. Secondly, cough, sputum production and sputum purulence are the key symptoms of bronchiectasis but comprise only one-third of the weight of the respiratory symptom domain of the QOL-B questionnaire. A questionnaire that gives greater weight to these symptoms may be more relevant to inhaled antibiotic treatment. The difference in weighting as well as recall period may explain differences observed in the RESPIRE trials, where the St George's Respiratory Questionnaire improved with treatment, but the QOL-B questionnaire did not [20, 21]. Indeed, the widespread use of inhaled antibiotics in clinical practice suggests that physicians and patients find them beneficial and that improvements in these three key symptoms are considered clinically important by patients [30].
Our post hoc analysis of quality-of-life data has provided a more detailed insight to the true drug response experienced in the AIR-BX trials and may provide an improved understanding of the phenotype of the ideal patients to be recruited into antibiotic trials. Nevertheless, our study has important limitations. Only 81% of the original trial cohort was available for re-analysis, although we have shown previously that the cohorts of included/excluded patients had no significant differences [18]. Our post hoc analysis is exploratory by definition and requires confirmation in further cohorts. In particular, we tested only one formulation of inhaled antibiotics in the form of aztreonam and validation with other antibiotics would be of interest. The effect of inhaled antibiotics on cough, sputum production and sputum purulence is mediated by the ability to reduce bacterial load, and the recent study by Laska et al. [9] of 16 inhaled antibiotics trials found no heterogeneity between aztreonam (pooled bacterial load reduction 2.6 log units, 95% CI 2.1–3.1) and the other antibiotics included (pooled reduction 2.3 units, 95% CI 1.2–3.4) in both their ability to reduce bacterial load and to improve symptoms (heterogeneity I2=1%). This strengthens the view that these results will be generalisable. Only the central conclusion of the study, that inhaled antibiotics improve cough, sputum production and sputum colour remained statistically significant after adjustment for multiple comparisons. Therefore, the results showing that baseline symptoms predict treatment response should be treated with caution as given the number of comparisons performed, some of the results may have arisen by chance. However, these are not entirely independent analyses, and the number, consistency and biological plausibility of our findings give confidence in their overall pattern. In the original AIR-BX studies a symptomatic benefit was seen in the first 4-week treatment period which was not evident during the second on-treatment period [17]. The reasons for this are unknown, but they also impacted on our analysis where the strength of associations with symptoms were strongest in the first 4 weeks and weaker when analysed over the full study duration. No specific minimum clinically important difference has been established for individual symptoms on the 4-point scale, and so while we found statistically significant differences favouring aztreonam versus placebo for cough, sputum purulence and sputum colour, further work is required to understand the clinical impact of these changes. Nevertheless, change of 1 point on the 4-point scale is substantial, as it indicates a change from “a lot” to “a moderate amount”, “a little” to “not at all”, “often” to “sometimes” or “sometimes” to “never” for individual symptoms, to give some examples. The absolute changes observed in our study would generate a number needed to treat for a cough and sputum production of between 3 and 5 patients to achieve a 1-point change in each symptom.
In conclusion, our results suggest aztreonam improves cough, sputum production and sputum colour, but does not significantly affect other symptoms in bronchiectasis. Inhaled antibiotic treatment may be most effective in patients with daily cough and producing discoloured sputum, and clinicians may wish to avoid treatment in patients with significant breathlessness and wheeze. Future trials should consider enrolling patients with a higher burden of bronchitic symptoms and develop symptom evaluation tools that give greater weight to these symptoms.
Supplementary material
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 material ERJ-00608-2020.Supplement
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-00608-2020.Shareable
Footnotes
This article has supplementary material available from erj.ersjournals.com
Support statement: The AIR-BX studies were funded by Gilead Sciences. This work was supported by the Innovative Medicines Initiative (IMI) and EFPIA companies under the European Commission funded project, iABC (grant 115721), the European Respiratory Society through the EMBARC2 consortium. EMBARC2 is supported by project partners Chiesi, Grifols, Insmed, Novartis and Zambon. J.D. Chalmers is supported by the GSK/British Lung Foundation Chair of Respiratory Research. Funding information for this article has been deposited with the Crossref Funder Registry.
Conflict of interest: M.L. Crichton has nothing to disclose.
Conflict of interest: M. Lonergan has nothing to disclose.
Conflict of interest: A.F. Barker has nothing to disclose.
Conflict of interest: O. Sibila has nothing to disclose.
Conflict of interest: P. Goeminne has nothing to disclose.
Conflict of interest: A. Shoemark has nothing to disclose.
Conflict of interest: J.D. Chalmers reports grants from Gilead, during the conduct of the study; grants and personal fees from AstraZeneca, GlaxoSmithKline, Grifols, Insmed and Boehringer Ingelheim, personal fees from Chiesi, grants from Gilead, outside the submitted work.
- Received January 31, 2020.
- Accepted March 31, 2020.
- Copyright ©ERS 2020