Roflumilast M2‐111.

Study characteristics
Methods	Study design: parallel‐group study Randomisation: randomised, double‐blind, placebo‐controlled trial Trial duration: 52 weeks Intention‐to‐treat analysis: stated
Participants	Setting: M2‐111 was conducted at 188 centres in 6 countries, and M2‐112 at 159 centres in 14 countries Participants: 1176 participants were randomised in this study (roflumilast: 500 µg: 568; placebo: 608) Baseline characteristics: severe COPD according to GOLD criteria grades III and IV, mean age 64 to 65 years, 72% male Inclusion criteria: aged ≥ 40 years, post‐bronchodilator FEV₁ < 50% predicted, reversibility < 15%, mean post‐bronchodilator FEV₁ 42%, FEV₁/FVC ≤ 0.7 with smoking history > 10 pack‐years, 40% current smokers, 60% ex‐smokers, average 46 to 48 pack‐years Exclusion criteria: history of asthma, lung cancer, or bronchiectasis; need for long‐term oxygen therapy; known α₁‐antitrypsin deficiency, clinically significant cardiopulmonary comorbidity Total numbers of participant withdrawals: data combined with M2‐112 showing 433 (33%) and 348 (26%) from treatment and control groups, respectively
Interventions	Run‐in: 4 weeks with placebo Roflumilast 500 µg once daily Placebo once daily Concomitant medication Short‐acting anticholingeric: 891 patients on short‐acting anticholinergics SABA: salbutamol as rescue medication Corticosteroid: 943 patients continued corticosteroid use LABA: none Used alongside corticosteroids, anticholinergics, and rescue short‐acting β₂‐agonists 54% overall (available to all)
Outcomes	Primary outcomes: change from baseline to endpoint in post‐bronchodilator FEV₁; number of moderate or severe exacerbations per patient per year Secondary outcomes: change from baseline in SGRQ total score; change from baseline in pre‐bronchial FEV₁, post‐bronchodilator FEV in 6 seconds and in FVC; FEF rate between 25% and 75% vital capacity; number of moderate or severe COPD exacerbations requiring systemic corticosteroid treatment per patient per year
Notes	NCT00076089/BY217/M2‐111. Funded by AstraZeneca
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The randomisation sequence was generated by a multiplicative congruent pseudo‐random numbers generator programme (programme RANDOM, based on Fishman and Moore)
Allocation concealment (selection bias)	Low risk	"Each study participant who qualified was assigned a number in sequential order. Code labelling prevented the investigator and the patient from knowing which drug was administered"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	The trial was double‐blind
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assumed that this would be low risk; however, no available information
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data combined with M2‐112
Selective reporting (reporting bias)	Low risk	Trial registered at clinicaltrials.gov; outcomes reported as planned. M2‐111 and M2‐112 data combined
Other bias	Low risk	None