7. ROB table: tocilizumab vs standard care(SC)/placebo. Time to clinical improvement.

Study	1.Randomisation	2.Deviations from intervention	3.Missing outcome data	4.Measurement of the outcome	5.Selection of the reported results	Overall risk of bias
Gordon REMAP‐CAP 2021	Low	Some concerns[1]	Low	Some concerns[2]	Low	Some concerns
Hermine CORIMUNO‐19 2020	Low	Some concerns[3]	Low	Some concerns[4]	Low	Some concerns
Rosas COVACTA 2021	Low	Low	Low	Low	Low	Low
Salama EMPACTA 2020	Low	Some concerns[5]	Low	Low	Low	Some concerns
Stone 2020	Low	Low	Low	Low	Low	Low
Salvarani 2020	Low	Some concerns6	Low	Some concerns7	Some concerns8	Some concerns

¹ Quote: “open‐label” Comment: unblinded study. Deviations from intended intervention arising because of the study context: No participant cross‐over. Administration of co‐interventions of interest: antivirals (Remdesivir, 32.8%) and corticosteroids (> 80%) were administered, but numbers per group were not reported. Hence, this domain was rated as some concern as not enough information on deviations that arose because of the trial context were reported. Participants were analysed according to their randomised groups for the outcome. Of note, 13 vs 10 participants were excluded from the analysis post‐randomisation for reasons related to missing data. This method was considered appropriate to estimate the effect of assignment to intervention.

² Comment: method of measuring outcome probably appropriate. Measurement or ascertainment of outcome probably does not differ between groups. Unblinded study (outcome assessor). Assessment of this outcome requires clinical judgement and can be influenced by knowledge of the intervention assignment, but is not likely in the context of the pandemic.

³ Quote: “Open‐label study” Comment: unblinded study.

Deviations from intended intervention arising because of the study context: three participants in the treatment group did not receive study drug. Administration of co‐interventions of interest (antivirals, corticosteroids and biologics) were reported. The proportions of participants receiving antivirals and steroids were imbalanced between two arms (> 10% absolute difference between the two arms) for steroids. This deviation could affect the outcome and was not balanced. Nevertheless, this domain was rated as 'Some Concerns' as it is impossible to distinguish deviation because of trial context and deviation because of intervention effect.

Participants were analysed according to their randomized groups for the outcome. Of note, 1 vs 0 participants were excluded from the analysis because of consent withdrawal. Nevertheless, we consider the analysis appropriate to estimate the effect of assignment to intervention.

⁴ Comment: method of measuring the outcome probably appropriate.

Measurement or ascertainment of outcome probably does not differ between groups Unblinded study. Assessment of this outcome requires clinical judgement and can be influenced by knowledge of the intervention assignment, but is not likely in the context of the pandemic.

⁵ Quote: “double‐blind, placebo‐controlled trial.” “A site blinding plan was established at each site to identify which personnel would be blinded or unblinded at a site level. A pharmacy manual and specific training in addition to completion of a site blinding plan was provided to each site. Each site had an unblinded pharmacist that randomized the participant and prepared and labeled study medication in the same method for both tocilizumab and placebo. The remainder of the study team was blinded to treatment assignment. There was no communication during the study between unblinded and blinded members. In addition, there was an unblinded medical monitor available to answer questions from the unblinded site staff. Placebo was not provided and consisted of an unaltered saline infusion bag, the same as would be used to prepare tocilizumab. The volume of tocilizumab diluted in saline appears colorless and matches saline.” Comment: blinded study. Participants were blinded. Carers were probably blinded.

Participants were analysed according to their randomised groups for the outcome. Of note, 10 vs 1 participants were excluded from the analysis post‐randomisation because they did not receive the drug. This method was considered inappropriate to estimate the effect of assignment to intervention for this time‐to‐event outcome. There was probably no substantial impact of failure to analyse participants according to their randomised groups.

⁶ Quote: "the trial was open label" Comment: unblinded study.

Deviations from intended intervention arising because of the study context: cross over: 15 (23%) participants in the standard care arm received the study treatment. For 12 (18%) the studied treatment was administered because of clinical worsening as planned in the protocol. Nevertheless, this decision could have been influenced by the trial context. Administration of co‐interventions of interest were reported and not balanced: antivirals (35% vs 47%) and corticosteroids (10% vs 10.6%). These deviations could affect the outcome and were not balanced. Nevertheless, this domain was rated as 'Some Concern' as it is impossible to distinguish deviation because of trial context and deviation because of intervention effect. Data for the outcome were analyced using intention‐to‐treat analysis. This method was considered appropriate to estimate the effect of assignment to intervention.

⁷ Comment: method of measuring the outcome probably appropriate. Measurement or ascertainment of outcome probably does not differ between groups Unblinded study. Assessment of this outcome requires clinical judgement and can be influenced by knowledge of the intervention assignment, but is not likely in the context of the pandemic.

⁸ Comment: the protocol and statistical analysis plan were available. The outcome time to clinical improvement (defined as time to discharge) is not mentioned in the protocol or registry. No information on whether the results for these outcomes were selected from multiple outcome measurements or analyses of the data.