Bernstein 2017

Methods	A randomised, double‐blind, double‐dummy, parallel‐group, 24‐week study
Participants	Population: intent‐to‐treat population included 1504 adults and adolescents (aged > 12 years) with an asthma diagnosis for at least 12 weeks who were well controlled on ICS/LABA Baseline characteristics: 82% of participants were white; 64% female; mean age 43.5 years. At randomisation, participants had a mean per cent predicted FEV₁ of 90.24% Inclusion criteria: required to have FEV₁ ≥ 80% of predicted normal value, and to have received treatment with ICS/LABA (equivalent to fluticasone propionate and salmeterol 250/50 twice daily), either as a fixed‐dose combination or through separate inhalers, for at least 12 weeks. Patients had to be able to replace their current SABA with albuterol/salbutamol Exclusion criteria: history of life‐threatening asthma in previous 5 years; evidence of concurrent respiratory disease or other clinically significant medical condition; ongoing respiratory infection within previous 4 weeks; use of tobacco products within previous 3 months or historical use ≥ 10 pack‐years; severe milk protein allergy or specific drug allergy; asthma exacerbation that required oral corticosteroids within previous 12 weeks, or that resulted in overnight hospitalisation requiring additional asthma treatment within previous 6 months
Interventions	• Fluticasone furoate and vilanterol 100/25 μg once daily (this arm of the study was not used in the review) • Fluticasone propionate and salmeterol 250/50 μg twice daily • Fluticasone propionate 250 μg twice daily
Outcomes	Primary outcome: change from baseline in evening trough FEV₁; safety was also assessed
Notes	Sponsored by GSK
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Eligible patients were randomised 1:1:1 via an interactive voice response system to receive 1 of 3 blinded study treatments
Allocation concealment (selection bias)	Unclear risk	Methods used for allocation concealment were not clearly reported
Blinding (performance bias and detection bias) All outcomes	Low risk	Study authors reported that it was carried out in a double‐blind double‐dummy manner, and described how this was achieved. Therefore it is unlikely that blinding was broken
Independent Assessment of causation (detection bias) Asthma‐related events	High risk	No report of independent assessment of SAEs carried out
Incomplete outcome data (attrition bias) All outcomes	Low risk	Reasons for attrition were given by study authors. Moreover the proportion of withdrawals, in both arms of the study, is relatively small: more than 80% of participants completed the study
Selective reporting (reporting bias)	Low risk	Data were extracted for all outcomes of the review