Blum 2015.

Methods	Study design: Randomised, double‐blind, placebo‐controlled trial Study duration: December 2009 to May 2014
Participants	Inclusion criteria Setting: 7 centres Country: Switzerland Health status: inpatient adults with community‐acquired pneumonia Number: treatment (402); control (400). Total randomised = 802 participants Median age (IQR): Treatment: 74 years (61 to 83 years) Control: 73 years (61 to 82 years) Exclusion criteria Permanent inability for informed consent Active IV drug use Acute burn injury GI bleeding within the past 3 months Known adrenal insufficiency A condition requiring more than 0.5 mg/kg per day prednisone equivalent Pregnancy or breastfeeding Severe immunosuppression Cystic fibrosis Active TB Antibiotic therapy: according to ERS/ESCMID guidelines
Interventions	Intervention: PO prednisone 50 mg x 1/d versus placebo Day start: within 24 hours from arrival to hospital Planned duration: 7 days Follow‐up: 30 days
Outcomes	Primary outcome: time to clinical stability defined as time (days) until stable vital signs for 24 hours Mortality outcome definition: 30‐day all‐cause mortality Other relevant outcomes: Clinical failure: defined as number of participants not reaching clinical stability on day 5 Need for invasive MV Need for ICU transfer Length of ICU stay Time to clinical cure: defined as time to stable vital signs for 24 hours or longer Length of hospitalisation Pneumonia complications: defined as empyema rates Superinfections Adverse events: any adverse event, GI bleeding, hyperglycaemia, neuropsychiatric adverse events, and cardiac adverse events Full agreement between outcomes in registry vs results: only primary outcome defined in registry Full agreement between outcomes in methods vs results: yes, except for timing of CAP scores defined only in results
Notes	Funding source: academic We contacted study authors for additional information, who provided missing data regarding mortality rates for the subgroups of severe pneumonia and COPD and mean values for length of hospital stay.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated
Allocation concealment (selection bias)	Low risk	Allocation was concealed with a prespecified computer‐generated randomisation list kept centrally at the pharmacy of the main study centre.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Outcome assessor was blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat data presented.
Selective reporting (reporting bias)	High risk	Only primary outcome reported in registry, differences in outcome description between methods and results.
Other bias	Low risk	Sample size calculation without early stop