| Study |
Bias |
| Randomisation process |
Deviations from intended interventions |
Missing outcome data |
Measurement of the outcome |
Selection of the reported results |
Overall |
| Authors' judgement |
Support for judgement |
Authors' judgement |
Support for judgement |
Authors' judgement |
Support for judgement |
Authors' judgement |
Support for judgement |
Authors' judgement |
Support for judgement |
Authors' judgement |
Support for judgement |
| Beigel 2020 |
Low risk of bias |
Randomisation was stratified by study site and disease severity in a 1:1 ratio and the allocation sequence was probably concealed. There were no baseline imbalances that would suggest a problem with randomisation. |
Some concerns |
Participants and care takers were adequately blinded. The analysis was inappropriate (as‐treated population). The deviation from an appropriate analysis population is small so that it probably had no substantial impact on the outcome results. |
Some concerns |
There is an uncertain amount of missings due to competing risk of death. It is uncertain, whether the result is biased or not. |
Low risk of bias |
Measuring of the outcome was appropriate. The outcome was assessed using standardised methods and assessors were unaware of the treatment assignments. |
Low risk of bias |
The data was analysed in accordance with a pre‐specified protocol. |
Some concerns |
Overall judged some concerns due to inappropriate analyses and non‐plausible participant selection. The method of the randomisation process was not provided, but there were no baseline imbalances that would suggest a problem with randomisation. Blinding was appropriate, and outcome measurement was according to a pre‐specified protocol. |
| Spinner 2020 |
Low risk of bias |
Block‐randomisation was not stratified and web‐based. Sites did not have access to the randomisation list. There were no baseline imbalances that would suggest a problem with randomisation. |
Low risk of bias |
Participants and care takers were aware of the assigned intervention. There was a neglectable number of participants with deviations from the intended intervention and the analysis was appropriate (intention‐to‐treat population). |
Some concerns |
There is an uncertain amount of missing data due to competing risk of death. It is uncertain, whether the result is biassed or not. |
Some concerns |
Outcome assessors were aware of the treatment assignments. Knowledge of intervention received could have affected outcome measurement. |
Low risk of bias |
Data was reported according to a study protocol providing details of time points, analyses and methods for categorisation and registration of adverse events. |
Some concerns |
Overall judged some concerns for risk of bias. Details about randomisation, blinding of outcome assessors and a protocol were provided. It was an open‐label study, and knowledge of intervention received could have affected outcome measurement to some degree. There is an uncertain amount of missing data due to competing risk of death. |
| Wang 2020 |
Some concerns |
Randomisation was carried out 2:1 and in stratified blocks. Concealed envelopes were prepared. There were baseline differences in gender distribution, respiratory status at baseline, co‐morbidities and time from symptom onset suggesting possible randomisation problems. |
Low risk of bias |
The study was carried out double‐blinded and the outcome was assessed by safety set. |
Some concerns |
There is an uncertain amount of missing data due to competing risk of death. It is uncertain, whether the result is biassed or not. |
Low risk of bias |
Outcome assessors were unaware of the treatment assignments and could not have been affected by knowledge of intervention. |
Low risk of bias |
The trial protocol provided details of time points and analysis. The statistical analysis plan could not be found. |
Some concerns |
Overall judged some concerns for risk of bias. Details about randomisation, blinding of outcome assessors and a protocol were provided. There were baseline differences between intervention and control group suggesting a problem with block wise randomisation process. There is an uncertain amount of missing data due to competing risk of death. |
| WHO Solidarity France 2021 |
Low risk of bias |
Randomisation appears appropriate with concealed allocation. There is an underrepresentation of female participants. |
Low risk of bias |
The study was open‐label and deviation from intended intervention could be because of experimental context. However, the small amount of deviation could be random, and an appropriate analysis was used. |
Low risk of bias |
Data was available for all participants. |
Some concerns |
Recording and assessment of seriousness of adverse events in an unblinded trial could be influenced by the knowledge of treatment group by the assessors, but direction of bias remains unclear and high number of trial centres with many different assessors should rule out systematic bias. |
Some concerns |
The outcome was reported additionally to a pre‐defined outcome set. Bias in result selection seems unlikely, since the reported outcome is common in safety sets and does not change the safety profile. |
Some concerns |
Overall judged some concerns. Randomisation was appropriate. The open‐label design could have influenced the assessment and could have caused deviations in the intended intervention. The outcome was reported additionally to a pre‐defined outcome set but selection bias. |
