The inclusion criteria were: male or female patients ≥40 years old; an established COPD clinical history ; a post-salbutamol FEV1/forced vital capacity (FVC) ratio <0.70 and a post-salbutamol FEV1 of ≥30 % and ≤70 % of predicted normal values; a dyspnoea score of ≥2 (modified Medical Research Council [mMRC] Dyspnoea Scale); current or former (stopped smoking for ≥6 months) cigarette smokers with a history of cigarette smoking of ≥10 pack-years. Key exclusion criteria were: asthma/other respiratory disorders; hospitalisation for pneumonia within 12 weeks of screening; a documented history of ≥1 COPD exacerbation requiring oral corticosteroids, antibiotics and/or hospitalisation in the 12 months preceding screening.
All patients provided written, informed consent prior to conducting any study-specific procedures. This study was approved by local ethics committees (Additional file 1) and performed in accordance with the Declaration of Helsinki  and Good Clinical Practice guidelines .
Study design, randomisation and treatment
This study (Additional file 2) was a phase IIIb, multicentre, randomised, double-blind, double-dummy, parallel-group trial (GSK study number DB2116134; www.clinicaltrials.gov registration number NCT01822899) conducted in 69 centres in eight countries (Czech Republic, Denmark, Germany, Hungary, The Netherlands, Poland, Russian Federation and Spain) between 2 April and 7 October 2013.
A validated computer system (RandAll; GSK, Brentford, UK) was used to generate a central randomisation schedule. Patients were randomised, using a Registration And Medication Ordering System (RAMOS; GSK, Brentford, UK), 1:1 to receive either UMEC/VI or FP/SAL. Patients and study personnel were blinded to the study medication.
After screening, eligible patients had a 7–14-day run-in period, in which as-needed salbutamol, mucolytics and as-needed oxygen therapy (≤12 h/day) were the only permitted COPD treatments, for assessment of baseline salbutamol use and disease stability. Following randomisation, patients received either UMEC/VI 62.5/25 mcg (delivered doses 55/22 mcg) via the ELLIPTA®Footnote 1 dry powder inhaler (DPI) once daily (morning) and placebo via the DISKUS®Footnote 2 inhaler (twice daily, morning and evening approximately 12 h apart) or FP/SAL 500/50 mcg via the DISKUS inhaler twice daily and placebo via the ELLIPTA DPI (once daily in the morning) for 12 weeks. There were further study visits at Weeks 4, 8 and 12 (end of treatment), and a 7 ± 2 day follow-up safety assessment. Patients were permitted to use salbutamol for as-needed symptom relief throughout the study, as long as it was withheld in the 4 h prior to spirometry testing. Further details of restricted and permitted concomitant COPD medications are provided in the Additional file 3. Treatment compliance was assessed by reviewing the inhaler dose counters at each study visit.
Efficacy (lung function) assessments
Spirometry was conducted at each visit. Baseline spirometry assessments were recorded prior to randomisation, during the same study visit. The primary endpoint was change from baseline in weighted mean (wm) FEV1 over 0–24 h on Day 84, calculated from pre-dose FEV1 and post-dose FEV1 evaluations at 5 and 15 min and 1, 3, 6, 9, 12 (pre-evening dose), 13, 15, 18, 23 and 24 h after the morning dose. The secondary endpoint was change from baseline in trough FEV1 on Day 85 (i.e., the mean of the FEV1 values recorded 23 h and 24 h after morning dosing on Day 84). Other lung function endpoints included (change from baseline unless otherwise stated): peak FEV1 over 0–6 h post-dose on Days 1 and 84; time to onset (an increase of ≥0.100 L above baseline in FEV1 during 0–6 h post-dose on Day 1); proportion of patients achieving an increase in FEV1 ≥ 12 % and ≥0.200 L above baseline at any time during 0–6 h post-dose on Day 1; wmFVC 0–24 h post-dose on Day 84; trough FVC on Day 85; and wmFVC 0–6 h post-dose on Days 1 and 84. The proportion of patients achieving an increase in FEV1 ≥ 0.100 L above baseline on Day 1 at 5 and 15 min, and 1, 3 and 6 h post-dose was evaluated in a post hoc analysis.
Symptomatic endpoints and health outcomes
Patients completed daily diaries, including rescue medication use (puffs/day, percentage of rescue-free days were calculated). Dyspnoea was assessed using the Baseline Dyspnoea Index (BDI) focal score at baseline, and the Transition Dyspnoea Index (TDI) focal score on Days 28, 56 and 84. Quality of life was assessed using the St George’s Respiratory Questionnaire for patients with COPD (SGRQ-C) at baseline and on Days 28 and 84. Health outcome assessments were evaluated using the EuroQol-5D (EQ-5D) questionnaire at randomisation and on Day 84. The COPD Assessment Test (CAT) was used to assess COPD-related health status at baseline and on Day 84.
Safety and tolerability included monitoring adverse events (AEs) throughout the study. AEs were coded using the Medical Dictionary for Regulatory Activities. COPD exacerbations were recorded. Vital signs were evaluated on Days 1 and 84.
The sample size calculation was based on a two-sided 5 % significance level and an estimated residual standard deviation of 0.220 L for wmFEV1 based on a mixed model for repeated measures (MMRM) analyses of previous studies in patients with COPD [6–8, 13]. Two hundred and eighty-four patients/group would have 90 % power to detect a 0.060 L treatment difference in 0–24 h wmFEV1. Assuming a 20 % drop-out rate, approximately 710 patients (355/group) were to be randomised.
All analyses were conducted on the intent-to-treat population (all randomised patients who took at least one dose of study medication). To account for multiplicity across endpoints, a step-down, closed-testing procedure was used. If the primary endpoint was statistically significant at the 5 % level, then the secondary endpoint was evaluated. If the latter was also statistically significant (5 % level) then inferences at the 5 % significance level would be made for all other comparisons.
An analysis of covariance (ANCOVA) model (covariates: baseline FEV1, smoking status and treatment) was used to analyse the 0–24 h wmFEV1 on Day 84. Trough FEV1 on Day 85 was analysed using MMRM analysis with covariates of baseline FEV1, smoking status, day, treatment, day by baseline interaction and day by treatment interaction, where day is nominal. The primary and secondary endpoints were also descriptively analysed by using the FEV1 % predicted to categorise patients as GOLD B (FEV1 ≥ 50 % predicted) or GOLD D (FEV1 < 50 % predicted), as all patients were required to have an mMRC score ≥2.