Patients meeting the inclusion/exclusion criteria at screening entered a 7–14-day run-in period during which long acting bronchodilator therapy was washed out; inhaled corticosteroids were continued at the same dose from ≥4 weeks prior to screening and throughout the study (Additional file 1: Figure S1). After the run-in, patients were randomised equally to five treatment groups, to inhale twice daily nebulised ensifentrine 0.75, 1.5, 3 or 6 mg or placebo. Patients then attended the study centre weekly for 4 weeks. Rescue salbutamol was permitted, but not within 8 h before spirometry assessments.
The study was approved by independent ethics committees at each institution, and was performed in accordance with the Declaration of Helsinki and Good Clinical Practice (ICH/CPMP/135/95). The study is registered at ClinicalTrials.gov (NCT03443414) and EudraCT (2016–005205-40).
Eligible patients were: males or females aged 40–75 years, inclusive; diagnosed with COPD for at least 1 year; clinically stable for at least 4 weeks; post-bronchodilator forced expiratory volume in 1 s (FEV1) 40–80% predicted normal and FEV1/forced vital capacity ratio ≤ 0.7; and current or former smokers with smoking history of at least 10 pack-years. To remain eligible, pre-dose FEV1 at randomisation was to be ±20% of the screening value. All patients provided written informed consent prior to any study-related procedure. Full inclusion and exclusion criteria are in the additional file.
At the randomisation visit, baseline (pre-dose) data were collected for spirometry (FEV1), St. George’s Respiratory Questionnaire – COPD Specific (SGRQ-C), Baseline Dyspnoea Index, and Medical Research Council dyspnoea scale (MRC). Spirometry was assessed pre-dose and up to 3 h post-dose on Weeks 1, 2 and 3, and up to 12 h post-dose on Day 1 and Week 4. At Week 4, Transition Dyspnoea Index (TDI), SGRQ-C, MRC and patient’s global assessment of change were assessed pre-dose. The patient’s global assessment of change is a study-specific questionnaire in which patients were asked, “Compared with prior to the study start, how do you feel your breathing is?”, graded on a scale of 1 = much worse to 5 = much better, with 3 = no change. Daily throughout the study (including between screening and randomisation) patients used an e-diary to record rescue medication use and COPD symptoms (using the Evaluating Respiratory Symptoms [E-RS™:COPD] questionnaire). Vital signs (blood pressure and pulse rate) and 12-lead electrocardiograms were assessed pre- and post-dose on all visits, with Holter monitor data collected over the 24-h periods prior to randomisation (for baseline) and the Week 4 visit. Adverse events were captured over the study duration.
The primary objective was to investigate the placebo-corrected effect of ensifentrine on change from baseline in peak FEV1 (assessed over 0–3 h) on Week 4. Secondary efficacy variables were: morning trough FEV1 (measured 15 min pre-dose) on Weeks 1–4; average FEV1 over 0–12 h on Day 1 and Week 4; average FEV1 over 0–3 h on Day 1 and Weeks 1–4; peak FEV1 on Day 1 and Weeks 1–3; E-RS™:COPD total score and average daily use of rescue medication over Weeks 1, 2, 3 and 4; and TDI, SGRQ-C total score (mean and percentage of responders), MRC and patient’s global assessment of change after 4 weeks. Ensifentrine safety was assessed by adverse events, electrocardiogram data, Holter monitoring, laboratory safety tests and vital signs.
Randomisation and masking
Patients were assigned to treatment groups in accordance with a randomisation list generated by the sponsor’s contract research organisation. Patients, investigators, site staff and sponsor personnel were masked to treatment assignment for the duration of the study, with the exceptions of the sponsor’s clinical supply chain lead, and of designated unblinded personnel at the sites, who, because the active and placebo treatments did not visually match, administered the in-clinic dose to patients and were not to take part in any of the study assessments.
Peak FEV1 standard deviation was estimated to be 250 mL. With a two-sided test at a 5% significance level and 80 evaluable patients per group, it was estimated that there would be 80% power to detect a true difference of 111 mL between any two treatments. This detectable limit was considered sufficient to identify a minimal effective dose of ensifentrine. Thus 80 patients per group were to be randomised.
The primary endpoint (Week 4 peak FEV1) was analysed using a restricted maximum likelihood-based mixed model for repeated measured (MMRM), including fixed effects for treatment, visit and treatment by visit interaction, patient as random effect, baseline value as covariate and covariance structure by visit. Ensifentrine–placebo differences with 95% confidence intervals and corresponding two-sided p-values were calculated. To control for the familywise error rate, a fixed-sequence testing strategy was employed, with the highest ensifentrine dose tested vs placebo. If a statistically significant difference was found at the two-sided α level of 5%, the testing proceeded with the next lower dose. If a test was non-significant, testing stopped and the remaining null hypotheses accepted.
A similar mixed model for repeated measured (MMRM) method was used to analyse most of the secondary efficacy endpoints, with the same hierarchical testing within endpoint, although endpoints were tested independently. The average FEV1 endpoints were calculated using the linear trapezoidal method as the area under the curve divided by the length of the time interval of interest. E-RS™:COPD total score data and rescue medication use were averaged to give weekly scores. The percentage of SGRQ-C responders, defined as patients with an improvement from baseline of ≥4 units, was analysed using a logistic regression model adjusting for treatment and country and giving the odds for being a responder as outcome, treatment differences expressed as the odds ratio.
The efficacy data were analysed in the full analysis set, which comprised all randomised patients who received at least one dose of study medication and had at least one post-treatment efficacy assessment. Safety analyses were in the safety population, which was all randomised patients who received at least one dose of study medication.