We carried out a multicentre observational study of historical cohorts of COPD patients. Inclusion criteria were: age >35 years, smoking habits >10 pack-years, post-bronchodilator forced expiratory volume in 1s (FEV1)/forced vital capacity (FVC) <0.7, and the administration of least one dose of any inhaled antibiotic in the five years prior to their inclusion in the study. We did not establish any exclusion criteria, but for the effectiveness analysis, only those patients for whom exacerbation data were available for at least 1 year were included.

All members of the Spanish Society of Pneumology and Thoracic Surgery (SEPAR) were informed of the details of the study and were invited to take part in it, by means of an e-mail communication. Those centres interested in participating contributed applicable cases, but no other pre-determined criteria were used to select them. The study was approved by the Clinical Research Ethics Committee of Catalonia (number: CEI 18/94). Informed consent was not requested as information obtained from the medical history was dissociated from the personal identification data, in accordance with the current data protection law (Regulation (EU) No. 2016/679 of the European Parliament and of the Council of 27 April 2016 on Data Protection). The study follows the STROBE guidelines and conforms to the principles outlined in the Helsinki Declaration.

Patients who had received at least one dose of any inhaled antibiotic between 2013 and 2018 were identified using the hospital pharmacy records. This list was cross-referenced to information obtained from medical records on the diagnosis of COPD. Data were collected by the end of 2019, so that there would be at least 1 year of follow-up in all patients after starting to take inhaled antibiotics.

The data were collected via a standardized protocol. They included exacerbations, spirometry, radiological and microbiological data and treatment with inhaled antibiotics (active substance, dose, duration and adverse effects). Comorbidities were calculated by the Charlson comorbidity index.23

The primary outcome of the study was the difference in the mean annual COPD exacerbation rate in the year before and the year after starting inhaled antibiotics. COPD exacerbations were defined as acute episodes of sustained worsening of respiratory symptoms that required outpatient pharmacological treatment or hospital admission.21 Secondary outcomes were COPD symptoms (expectoration, sputum purulence and dyspnoea), microbiological profile and PPM eradication from sputum.

Sputum staining was analyzed following the validated scale of Murray et al.24 Bacteria isolated in sputum cultures were classified as PPM strains in the case of HI, Streptococcus pneumoniae, Moraxella catarrhalis, Staphylococcus aureus, methicillin-resistant S. aureus (MRSA), PA and other Gram-negative bacilli.

We defined CBI as the presence of at least three positive sputum cultures for the same PPM in a 12-month period, one month apart.25 Conversely, patients were considered free from chronic infection when at least three consecutive cultures were negative for the same PPM in a 12-month period, one month apart.25 Bronchiectasis were diagnosed by high-resolution computed tomography (HRCT).12

For both the descriptive and safety analysis, data from all the patients who had received at least one dose of inhaled antibiotic in the last five years were included. For the effectiveness analysis (primary and secondary outcomes), we only considered patients with a known number of exacerbations that appeared at least 1 year after the start of treatment (≥1 year of follow-up data). Moreover, a sensitivity analysis was carried out in those patients who used inhaled antibiotics for a minimum of 3 months after their prescription and in those who had bronchiectasis confirmed by HRCT.

Quantitative variables were expressed by the mean and standard deviation or the median and interquartile range, depending on whether they followed a normal distribution determined by the Kolmogorov–Smirnov test. If so, the Student's t-test was used for independent or paired means, while the Mann–Whitney or Wilcoxon U tests were applied to variables with non-normal distribution, respectively. Either the ANOVA or Kruskal–Wallis tests were used to compare more than two means or medians, depending on the specific case. Qualitative variables were expressed according to their relative percentage and compared with the Chi-square and Fisher's exact tests.

Variables associated with annual changes in the number of exacerbations before and after inhaled antibiotic treatment were assessed using a mixed-effects linear regression model allowing for random intercepts. The variables introduced into the multivariate model were selected on the basis of their clinical relevance: age, gender, presence of bronchiectasis, CBI by any PPM, post-bronchodilator FEV1 (%predicted), smoking habit and annual number of exacerbations in the two years prior to treatment with inhaled antibiotics.10 Statistical significance was defined by P<.05. The IBM® Statistical Package for the Social Sciences (SPSS®) software version 24 was used.

A total number of 709 patients from 35 hospitals throughout Spain were considered for the study. Of these, 693 (97.7%) met all the inclusion criteria for the descriptive and safety analysis (Fig. 1). Table 1 shows the baseline characteristics of the patients when inhaled antibiotics were prescribed. The mean age was 74.1 (standard deviation (SD)=8.5) years, and 86.3% of the study population were men. The mean FEV1 was 43.7% (SD=16.9%). Most of the patients presented with daily expectoration (76.9%), mainly mucopurulent (48%), while 94.4% and 88.9% were under treatment with long-acting beta-adrenoceptor agonist (LABA) or long-acting muscarinic antagonist (LAMA) bronchodilators, respectively, and 83.7%, with inhaled corticosteroids.

Fig. 1.

Study flowchart. BE: bronchiectasis; HRCT: high-resolution computed tomography.

(0.37MB).

The median number of sputum cultures collected per patient during the two years prior to the start of inhaled antibiotics was 6 (IQR=4–9) (range=1–54). The median time elapsed between the first positive sputum and the start of the treatment was 10.1 months (IQR=3.7–19) (range=0–84). At least one PPM was isolated in 100% of the patients, with PA being the most frequently involved (91.6%); 51.5% of patients met the criteria for CBI by any PPM, with PA again being the most frequent (Table 1). The main reason for starting inhaled antibiotics was the presence of frequent exacerbations in patients with positive cultures, regardless of the presence of criteria for CBI. In fact, when patients with and without these criteria were compared, there was no significant difference in the number of exacerbations (3.5 vs. 3.1; P=.202) or hospitalizations (1.7 vs. 1.5; P=.062) during the previous year.

Patients were given colistimethate sodium (82.5%), tobramycin (15.3%), ceftazidime (1.6%), gentamicin (0.4%) or amikacin (0.1%) (e-Table 1) for a median time of 8.9 (interquartile range (IQR)=3.3–22) months (range=0.1–71.5).

When the patients for whom HRCT information was available (625 patients; 90.2%) were considered separately, we found that 462 (73.9%) of them had bronchiectasis (Fig. 1). Compared with patients without bronchiectasis, these patients presented with increased daily expectoration (P<.001), more frequent mucopurulent or purulent sputum (P=.035), a higher proportion of CBI by PA or any PPM (both P<.001) and an increased use of long-term mucolytics (P=.021) and oral cyclic antibiotics (P=.011) (Table 1).

The exclusion of patients with insufficient follow-up data left a subtotal of 443 (63.9%) individuals for analysis (Fig. 1). e-Table 2 shows the differences between these patients and the remaining population.

During the two years prior to starting inhaled antibiotics, the mean number of exacerbations and hospital admissions progressively increased, as well as the proportion of patients with exacerbator phenotype (Fig. 2). However, when comparing the first year prior to and the first year subsequent to the start of inhaled antibiotics (primary outcome), we observed a significant decrease in the mean number of exacerbations (−33.3%; P<.001) (Table 2 and Fig. 2), mainly due to the lower percentage of patients with three or more events after the treatment (Fig. 3). The number of hospital admissions and days of hospitalization also dropped (−33.3%; P<.001 and −26.2%; P=.003, respectively) (Table 2 and Fig. 2). When the analysis was extended to a two-year period before and after antibiotics, the reduction of exacerbations (−27.6%; P<.001) and hospitalizations (−25%; P=.002) was still significant (Table 2 and Fig. 2). Moreover, the significant decrease in the percentage of patients with exacerbator phenotype was observed up to two years after starting the treatment (Fig. 2). These findings were maintained when patients (N=387; 55.8%) who received inhaled antibiotics for at least 3 months were considered (e-Figure 1).

Fig. 2.

Evolution of patients with COPD under treatment with inhaled antibiotics: (A) number of exacerbations and hospital admissions and (B) proportion of patients with exacerbator phenotype. N=443. IA: inhaled antibiotics. §=P<.001 (compared to the year before IA); #=P=.42 (compared to two years before IA); ¥=P<.001 (compared to two years before IA).

(0.19MB).

Fig. 3.

Comparison of the number of exacerbations one year before and one year after inhaled antibiotics. N=443. IA: inhaled antibiotics.

(0.18MB).

Twenty-one patients (4.7%) received treatment for less than 1 month, mainly due to side-effects that led to discontinuation (in 17 cases). In these patients, the mean number of exacerbations, hospitalizations and days of hospitalization in the subsequent year was significantly higher than in those who completed more than 1 month of treatment.

The multiple linear regression analysis (Table 3) identified independent factors associated with a significant change in the number of exacerbations. CBI by any PPM (beta coefficient=−0.22; 95% confidence interval (CI)=−0.31, −0.55; P<.001) and at least three exacerbations in each of the previous two years (beta coefficient=−0.18; 95% CI=−1.3, −0.42; P<.001) were associated with a significant decrease in the number of exacerbations. Finally, since the vast majority of CBI seen in our patients were by PA, we performed an additional multivariable analysis changing CBI by any PPM for CBI by PA, and the results did not change (beta coefficient: −0.25; 95% CI: −0.34, −0.59; P<.001).

Regarding secondary outcomes, inhaled antibiotics were associated with a significant reduction in daily expectoration (80.8% vs. 55.7% of patients; P<.024) and mucopurulent or purulent sputum (73% vs. 36.9%; P<.001) after one year of treatment, although spirometry performance slightly worsened after one year of treatment (69.4% vs. 67.7% of mean predicted FVC and 44.8% vs. 43.7% of mean predicted FEV1) (Table 4).

The microbiological profiles of the sputum samples show that PA and HI were isolated to a lesser extent (91.2% of patients vs. 54%; P<.001 and 15.1% vs. 6.3%; P=.015, respectively) (Table 4 and e-Figure 2). We also found a reduction in the prevalence of CBI by PA (49.4% vs. 24.8%; P<.001), while PA was eradicated in 38.8% of COPD cases (Table 4).

We sought to further analyze the effectiveness of inhaled antibiotics depending on the presence or absence of bronchiectasis (Fig. 1). After stratifying the data, we did not find any relevant differences in treatment effectiveness after one year. Both subgroups of patients separately showed a significant reduction in the number of exacerbations (−34.4% in patients with bronchiectasis; P<.001 and −36.8% in patients without bronchiectasis; P<.001), admissions (−33.3%; P<.001 and −33.3%; P=.003) and days spent in hospital (−22.8%; P=.061 and −35.7%; P=.003) (Table 2). Clinical improvement continued until two years after the initiation of inhaled antibiotics in both these subgroups of COPD patients (Fig. 4).

Fig. 4.

Evolution of patients with and without bronchiectasis before and after the start of inhaled antibiotic treatment: (A) number of exacerbations (upper graph) and hospital admissions (lower graph) and (B) proportion of patients with exacerbator phenotype. N=405. BE: bronchiectasis; IA: inhaled antibiotics. §=P<.001; ¥=P=.003; €=P=.015; ∏=P=.002; ∑=P=.04 (compared to the year before IA); #=P=.015 (compared to two years before IA).

(0.24MB).

We found that PA eradication was achieved in a significantly higher proportion of patients with no criteria for CBI by PA, compared to those patients who already had CBI by PA (49.2% vs. 37.0%; P=.013). Furthermore, we found that the effectiveness of the treatment varied according to the achievement of eradication of airway PA. PA-free patients presented less exacerbations (1.7; P<.001 vs. 2.2; P<.001) and hospitalizations (0.7; P<.001 vs. 1.0; P<.001) per year compared with patients with persistent PA-positive cultures (Fig. 5). There was a smaller percentage of deaths (23.3% vs. 32.3%; P=.046) and exacerbator phenotype (61%; P<.001 vs. 71.4%; P=.007) after 2 years of follow-up when PA was successfully eradicated (e-Table 3 and Fig. 5). Additional sociodemographic and clinical baseline features of both subgroups of patients may be checked on e-Table 3.

Fig. 5.

Evolution of patients with or without successful eradication of PA (sputum isolates or CBI) before and after the start of inhaled antibiotic treatment: (A) number of exacerbations and hospital admissions and (B) proportion of patients with exacerbator phenotype. N=404. IA: inhaled antibiotics; PA: Pseudomonas aeruginosa. §=P<.001; €=P=.015; £=P=.006; ∏=P=.001; ∑=P=.007 (compared to the year before IA); #=P=.009; ¥=P<.001 (compared to two years before IA).

(0.24MB).

A total of 176 patients (25.4%) suffered from adverse events related to treatment with inhaled antibiotics after a median time frame of 22.5 days (IQR=7.7–90). The most frequent side-effects were bronchospasms (10.5%), dyspnoea (8.8%) and cough (1.7%) (e-Table 4). Although none were classified as serious, medication had to be changed in 6.2% of patients and discontinued in another 15.9% after 56 days (IQR=17–122), mainly due to dyspnoea or cough (82% of cases).