The study is a multicenter, randomised, GCP monitored, controlled, open-label trial conducted in outpatients with COPD, non-CF bronchiectasis, or asthma with a P. aeruginosa-culture positive lower respiratory tract sample. The study was carried out at seven respiratory outpatient clinics in Denmark between October 2017, and March 2023. Outpatients with a P. aeruginosa-positive lower respiratory tract sample (sputum, tracheal secretion, bronchial secretion or bronchial alveolar lavage) obtained within the previous 30 days, regardless of prior P. aeruginosa-status, and with a physician-judged no need for hospitalisation, were systematically screened and consecutively invited to participate if they fulfilled inclusion criteria and no exclusion criteria (Additional file 1: study protocol). The study was approved by the Ethics Committees (H-15010949), the Danish Medicines Agency (EudraCT 2015–003399-58) and the Danish Data Protection Agency (HGH-2017–036), and was monitored by a national GCP unit. The trial is registered at ClinicalTrials.gov (NCT03262142). No financial incentive was provided to the investigators or participants.

Outpatients were randomly assigned 1:1 to either systemic dual antibiotic treatment (antibiotic group) or no antibiotic treatment (control group) and stratified by study site and age (≤ 70 years vs. > 70 years) (see Appendix for details regarding randomization sequence). The antibiotic intervention consisted of 14 days of combination therapy with piperacillin/tazobactam 4/0.5 g, administered intravenously four times daily, and oral ciprofloxacin 500 mg twice daily. Intravenous ceftazidime or meropenem was used if piperacillin/tazobactam could not be used because of allergy or antibiotic resistance.

Outpatients were screened based on results from routine microbiological examinations of the lower respiratory tract samples obtained from patients attending the outpatient clinics of the participating respiratory departments. Samples were ordered by clinical staff based on clinical symptoms of exacerbation of the underlying lung disease. Fever, fatigue, peripheral oxygen saturation, and tachypnoea at rest were used as parameters to guide the staff when assessing the need for hospitalisation. Antibiotics were administered in-hospital, since home-treatment with intravenous antibiotics was not available for all study sites at the time of implementation. However, between 2020 and 2022, one site (initiated in 2017), was able to provide treatment at home. We allowed a delay of initiating antibiotic treatment for up to six days in initiating therapy, since there, per the eligibility criteria, was no clinical indication for admission. Baseline measurements were obtained on the calendar date of recruitment (day 1) and follow-up visits were scheduled on day 14, 30, 60, 90, and 365. COPD assessment test (CAT), body mass index (BMI), Medical Research Council (MRC) dyspnoea score, and spirometry were assessed at all visits. Blood samples were drawn at day 1 (baseline) and day 14, and the outpatients underwent a high-resolution CT of the lungs at day 14 assess radiological signs of bronchiectasis at baseline.

The primary outcome was time to prednisolone and/or antibiotic requiring exacerbation, in a primary or secondary health care sector, or death from day 20 to day 365 from randomisation. Death was incorporated in the primary outcome to avoid lead-time bias as the death rate is expected to be high in this population of outpatients with severe pulmonary disease and could thus be incorrectly interpreted as protective of exacerbation. We chose to register events after 20 days from randomisation to avoid misclassifying the study intervention as a fulfilment of the primary outcome. A co-primary outcome of "days alive and out of hospital within 365 days" was degraded to the first secondary outcome by the trial statistician in agreement with the trial leadership (JE and JUSJ) since this outcome would be severely underpowered because of the premature closure of the trial (see " Statistical analysis"). This was done before the database was unblinded to the analysis (see Additional file 1).

The secondary outcomes were: 1) days alive and without hospitalisation from day 20 to day 365 from randomisation, 2) death within 365 days from randomisation, 3) number of admissions with exacerbation within 365 days from randomisation (defined as referral to emergency room or hospitalisation [9]), 4) number of days with non-invasive ventilation or invasive ventilation within 90 days from randomisation, 5) microbiological cure at day 90 (defined as P. aeruginosa-negative sputum culture until day 90; no microbiological cure was defined as a P. aeruginosa-positive sputum culture before or at day 90), 6) clinical cure at day 14 (defined as improvement of clinical signs and symptoms related to P. aeruginosa before or on day 14; clinical failure was defined as persistent or worsening of clinical signs and symptoms related to P. aeruginosa before or on day 14), 7) change in CAT score from randomisation to day 90, 8) change in BMI from randomisation to day 90, 9) change in forced expiratory volume in the first second (FEV1) from randomisation to day 90, and 10) decrease of ≥ 200 ml in FEV1 from randomisation to day 365.

The sample size was calculated using a group-sequential design, allowing for one interim analysis at half target recruitment, with a power of 80% to avoid type II error at a two-sided 5% significance level. Based on estimates and indicative figures in previous literature, a total of 150 patients (75 patients in each group) were required for the trial (see Additional file 1 for details) [3, 10,11,12].

Data were analysed using intention-to-treat (ITT) principles, including all available data, regardless of whether the participant received the intervention. The primary outcome was also analysed using a modified ITT analysis (in study participants who started but did not complete the intervention) and per protocol analysis (in study participants who completed the entire intervention). Completion of the intervention was defined as 14 days of antibiotic treatment in the dual systemic anti-pseudomonal antibiotic study group, and as no anti-pseudomonal treatment within 14 days from randomisation in the control group. Partial completion to intervention was defined as 1–13 days of antibiotic treatment in the antibiotic study group and ≥ 1 day of P. aeruginosa-active antibiotic treatment within 14 days from randomisation in the control group.

Data for the primary outcome analyses was analysed using an unadjusted Cox proportional hazard model and results reported as hazard ratio [HR] and 95% confidence limits. A Kaplan–Meier plot was used to describe the process of exacerbations and death in the study groups. A multivariable Cox proportional hazards model adjusting for sex (male vs. female), CAT-score (< 21 vs. ≥ 21) and FEV1% predicted (< 50% vs. ≥ 50%) at randomisation was also conducted. Secondary outcomes were compared between the study groups using t-test or Mann–Whitney U test for continuous data and χ2-test for nominal data. Analysis of covariance (ANCOVA) was used to model the effect of the intervention on changes in the mean of continuous outcomes, adjusting for the baseline value. All analyses were done using the statistical software SAS (version 9.4) and R (version 3.4.3). Sample size calculation was done using StudySize 2.0 (Frölunda, Sweden).

The interim analysis was planned at half target recruitment (75 patients), with a focus on reporting data on the primary outcome, all-cause mortality at day 365, microbiological cure at day 14 and assessment of study's futility. An independent data and safety monitoring board (DSMB) was appointed to review the trial's safety, efficacy, and progression (see Additional file 1). Due to the slow recruitment rate during the COVID-19 pandemic, the trial's steering committee decided to stop further recruitment in February 2022, when approximately 1/3 of the planned outpatients had been enrolled in the study. This decision was tested with the DSMB, who endorsed it (see Additional file 1). Due to the considerable reduction of the study size, the primary outcome was conducted solely as a "time to event" analysis, and the "days alive and out of hospital" analysis was degraded from a co-primary outcome to the first secondary outcome. As previously mentioned, this decision was made before data was unblinded to the analysis (see " Outcomes"). Data analyses were performed by an analysis team (TWK and AJ), including a trial statistician (TWK), after the final data from the last outpatients last follow-up visit was entered and the database was locked. All analyses were done before breaking of the randomisation code. The study group was presented to the results and unblinded at a scheduled unblinding-meeting after the analyses were performed.