Note: The risk of bias by domain corresponds to the highest risk of bias among outcomes by domain.
The overall risk of bias corresponds to the overall highest risk of bias assessed among outcomes.
| Bias | Author's judgement | Support for judgement |
| Randomization |
Low |
Quote: "Participants were assigned by
means of a centralized computer program to each
intervention, starting with balanced assignment
for tocilizumab, sarilumab, or control, with actual
proportions dependent on the number of
interventions available at each site." (report)
“Allocation concealment will be maintained by using centralized randomization that is remote from study sites.” (trial registry) Comment: Allocation sequence random. Allocation concealed. Imbalances in baseline characteristics appear to be compatible with chance. |
| Deviations from intervention |
Some concerns |
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: corticosteroids (252/272 vs 46/48 vs 293/312) were administered and balanced; antivirals (remdesivir: 107/341 vs 21/48 vs 133/389) were reported and imbalanced. This deviation was not balanced and could affect the outcome. 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 binary outcomes were analyzed using intention-to-treat analysis. This method was considered appropriate to estimate the effect of assignment to intervention. Participants were analyzed according to their randomized groups for the time-to-event outcomes. Of note, 13 vs 0 vs 10 participants were excluded from the analysis post-randomization for reasons related to missing data. This method was considered appropriate to estimate the effect of assignment to intervention. Risk assessed to be some concerns for outcome: Mortality (D28). Time to death. Time to clinical improvement. Serious adverse events. |
| Missing outcome data |
Low |
Comment: 826 patients randomized; 803 patients analyzed for serious adverse events, time to death, time to clinical improvement outcomes; 792 patients analyzed for mortality outcome.
Data available for all or nearly all participants. Risk assessed to be low for the outcomes: Mortality (D28). Time to death. Time to clinical improvement. Serious adverse events. |
| Measurement of the outcome |
Some concerns |
Comment: Method of measuring the outcome probably appropriate.
Measurement or ascertainment of outcome probably does not differ among groups. Unblinded study (outcome assessor) Mortality is an observer-reported outcome not involving judgement. Risk assessed to be low for the outcome: Mortality (D28). Time to death. Clinical improvement (defined as hospital discharge) requires clinical judgement and could be affected by knowledge of intervention receipt. Also, the authors reported on serious adverse events that may contain both clinically- and laboratory-detected events. All these outcomes can be influenced by knowledge of the intervention assignment, but is not likely in the context of the pandemic. Risk assessed to be some concerns for the outcomes: Time to clinical improvement. Serious adverse events. |
| Selection of the reported results |
Low |
Comment: The protocol, statistical analysis plan, and registry were available.
Mortality, time to clinical improvement and serious adverse events results were not selected from multiple outcome measurements or analyses of the data. Trial analyzed as pre-specified. Time to death results were probably not selected from multiple outcome measurements or analyses of the data. Trial probably analyzed as pre-specified with respect to time to death (all patients in ICU; pre-specified endpoint: ICU mortality censored at 90 days). Risk assessed to be low for the outcomes: Mortality (D28). Time to death. Time to clinical improvement. Serious adverse events. |
| Overall risk of bias |
Some concerns |