Objectives
Primary objective: To estimate the effect of corticosteroids compared with usual care or placebo on mortality up to 28 days after randomization. Secondary objectives: To examine whether the effect of corticosteroids compared with usual care or placebo on mortality up to 28 days after randomization varies between subgroups related to treatment characteristics, disease severity at the time of randomization, patient characteristics, or risk of bias. To examine the effect of corticosteroids compared with usual care or placebo on serious adverse events.

Study design
Prospective meta-analysis of randomized controlled trials. Both placebo-controlled and open-label trials are eligible.

Participants
Hospitalised, critically ill patients with suspected or confirmed COVID-19.

Intervention and comparator
Intervention groups will have received therapeutic doses of a steroid (dexamethasone, hydrocortisone or methylprednisolone) with IV or oral administration immediately after randomization.

The comparator groups will have received standard of care or usual care or placebo.

Main outcome
All-cause mortality up to 28 days after randomization.

Search methods
Systematic searching of clinicaltrials.gov, EudraCT, the WHO ISRCTN registry, and the Chinese clinical trials registry. Additionally, research and WHO networks will be asked for relevant trials.

Risk of bias assessments
These will be based on the Cochrane RoB 2 tool, and will use structured information provided by the trial investigators on a form designed for this prospective meta-analysis.

Summary of findings
We will use GRADE to assess the certainty of the evidence.

Statistical analyses
Trial investigators will provide data on the numbers of participants who did and did not experience each outcome according to intervention group, overall and in specified subgroups. We will conduct fixed-effect (primary analysis) and random-effects (Paule-Mandel estimate of heterogeneity and Hartung-Knapp adjustment) meta-analyses. We will quantify inconsistency in effects between trials using I2 statistics. Evidence for subgroup effects will be quantified by ratios of odds ratios comparing effects in the subgroups, and corresponding interaction p-values. Comparisons between subgroups defined by trial characteristics will be made using random-effects meta-regression. Comparisons between subgroups defined by patient characteristics will be made by estimating trial-specific ratios of odds ratios comparing intervention effects between subgroups then combining these using random-effects meta-analysis. Steroid interventions will be classified as high or low dose according to whether the dose is greater or less than or equal to 400 mg hydrocortisone per day or equivalent. We will use network meta-analysis methods to make comparisons between the effects of high and low dose steroid interventions (because one trial randomized participants to both low and high dose steroid arms).

PROSPERO registration number
CRD42020197242

Full protocol
The full protocol for this prospective meta-analysis is attached as an additional file, accessible from the Trials website (Additional file 1). To expedite dissemination of this material, the familiar formatting has been eliminated; this Letter serves as a summary of the key elements of the full protocol for the systematic review.