Goodacre 2013.

Methods	Double blind, randomised controlled trial. 34 emergency departments, UK.
Participants	Inclusion criteria: severe (BTS/SIGN quantified) asthma attack, age ≥16 years. Exclusion criteria: life‐threatening features, contraindication to MgSO₄, participant unable to give verbal/written consent, previous participation in the study; criteria amended to exclude those who had received MgSO₄ in the past 24 h. 1109 randomised. Intervention 1 (nebulised MgSO₄): 339 randomised. Mean age (years): 36.5 (14.8). Men:women: 107:232. Smokers: 98. Mean predicted PEF (L/min): 430 (118.8). Intervention 2 (intravenous MgSO₄): 406 randomised. Mean age (years): 35.6 (13.1). Men:women: 130:279. Smokers: 138. Mean predicted PEF (L/min): 431.8 (116.9). Control: 364 randomised. Mean age (years): 36.4 (14.1). Men:women: 112:252. Smokers: 127. Mean predicted PEF (L/min): 435.0 (110.8).
Interventions	Intervention 1: 100 mL 0.9% NaCl IV and 2 mmol MgSO₄ in 7.5 mL 0.9% NaCl nebulised. Intervention 2: 8 mmol MgSO₄ in 100 mL 0.9% NaCl IV and 7.5 mL 0.9% NaCl nebulised. Control: 100 mL 0.9% NaCl IV and 7.5 mL 0.9% NaCl nebulised. IV infusion given once over 20 mins, nebulisers given 3 times, each over 20 minutes.
Outcomes	Admission (4 h, 7 days); change in participant's assessment of breathlessness via visual analogue scale, change in PEF, heart rate, respiratory rate, BP, oxygen saturations (1, 2 h); adverse events (2 h); mortality, length of hospital stay, admission to HDU or ICU. Adverse events: treatment group 41 adverse events; control group 36 adverse events.
Notes	Funding: UK National Institute for Health Research Health Technology Assessment Programme.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Simple and blocked randomisation sequences used to allocate participants to numbered treatment packs.
Allocation concealment (selection bias)	Low risk	Allocated treatment pack numbers were only revealed after participant details recorded and the participant irreversibly entered into the trial.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants, hospital staff, and research staff were masked to allocated treatment.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participants, hospital staff, and research staff were masked to allocated treatment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Over 95% of randomised participants in each arm were included in primary analysis. All participants clearly accounted for in flow diagram. There was an inevitable 'drop off' in participants available at each time point for many of the secondary outcomes; it is unclear what impact this may have had on the results.
Selective reporting (reporting bias)	Low risk	Prospective trial registration identified. All primary and secondary outcomes listed in the trial register were reported.