Castro 2014a.

Study characteristics
Methods	Double‐blind, dose‐ranging RCT
Participants	606 participants with uncontrolled asthma randomised and 535 completed Main inclusion/exclusion criteria: 2‐6 exacerbations in the previous 12 months ACQ‐6 score ≥ 1.5 at least twice during screening morning pre‐bronchodilator FEV1 40%‐90% maintenance treatment with medium‐ to high‐dose ICS in combination with LABA for ≥ 12 months Mean age, years (SD): eosinophilic benralizumab 2 mg, 47 (12.8); eosinophilic benralizumab 20 mg, 47 (13.2); eosinophilic benralizumab 100 mg, 48 (12.9); eosinophilic placebo, 46 (11.7); non‐eosinophilic benralizumab 100 mg, 50 (11.5); non‐eosinophilic placebo, 50 (12.3) Male, n (%): eosinophilic benralizumab 2 mg, 23 (28%); eosinophilic benralizumab 20 mg, 33 (41%); eosinophilic benralizumab 100 mg, 22 (27%); eosinophilic placebo, 27 (33%); non‐eosinophilic benralizumab 100 mg, 42 (30%); non‐eosinophilic placebo, 42 (30%) Baseline mean (SD) FEV1 % predicted: eosinophilic benralizumab 2 mg, 65% (SD 15); eosinophilic benralizumab 20 mg, 64 (15); eosinophilic benralizumab 100 mg, 66 (16); eosinophilic placebo, 65 (15); non‐eosinophilic benralizumab 100 mg, 69 (15); non‐eosinophilic placebo, 67 (15) Allocation: eosinophilic benralizumab 2 mg, 81; eosinophilic benralizumab 20 mg, 81; eosinophilic benralizumab 100 mg, 80; eosinophilic placebo, 80; non‐eosinophilic benralizumab 100 mg, 140; non‐eosinophilic placebo, 142
Interventions	6 arms: benralizumab 2 mg or benralizumab 20 mg or benralizumab 100 mg or placebo delivered by 2 SC injections every 4 weeks for the first 3 doses (weeks 1, 4, and 8), then every 8 weeks (weeks 16, 24, 32, and 40)
Outcomes	Primary outcomes Annual exacerbation rate in eosinophilic participants Secondary outcomes in eosinophilic individuals Change from baseline, in FEV1 ACQ‐6 Overall symptom score AQLQ
Notes	52‐week multi‐national study with sites in 10 countries. The study protocol was developed by MedImmune and the corresponding author. The investigators collected and had full access to all study data, which were analysed by the funding source. The analysis was done solely by MedImmune; however, study authors helped determine which analyses were done and could request further ad‐hoc analyses. The report was written by the study authors with a medical writer funded by the funding source. The corresponding author had final responsibility for decision to submit for publication. Funding: MedImmune
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Interactive web/voice‐response system for random assignment
Allocation concealment (selection bias)	Low risk	Allocation concealment was ensured by the vendor systems and no study personnel or site had access to the system.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants, treating physicians, study investigators, and study statisticians were masked to treatment allocation.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	As above
Incomplete outcome data (attrition bias) All outcomes	Low risk	The withdrawal rates were even across groups.
Selective reporting (reporting bias)	Low risk	Results for most but not all listed primary and secondary outcomes were reported (e.g. symptoms score, AQLQ – shown in supplementary material in graphs only)