Barbaro 1998.

Methods	Double‐blind, multi‐centre, placebo‐controlled RCT. Cross‐over design: investigators crossed‐over participants who did not respond to intervention during first period (remained in Grade III or IVa coma) to alternative intervention.
Participants	527 participants with cirrhosis and overt hepatic encephalopathy (Grades III or IVa; Table 3), admitted to an intensive care unit. Diagnostic criteria corresponded to acute hepatic encephalopathy. Precipitating factors are described (Table 5). Mean ± SD age (grade III/IVa): flumazenil: 56 ± 11.5/53 ± 12 years; placebo: 48 ± 20/55 ± 13.5 years. Proportion of men: 69%. Aetiology of cirrhosis: alcohol 40%; hepatitis B/C 59%. Proportion testing positive for benzodiazepines at baseline (Table 6): 10/527 (1.9%) participants.
Interventions	Intervention comparison: intravenous infusion flumazenil 1 mg given over 3 to 5 minutes versus placebo (isotonic saline). Total dose of flumazenil: 1 mg. Cointerventions: lactulose 30 mL every 6 hours via nasogastric tube; antibiotics were given to 22 participants with sepsis in the flumazenil group and 8 in the placebo group
Outcomes	Outcomes included in meta‐analyses: mortality, hepatic encephalopathy (Table 2), and serious adverse events (Table 7) assessed for a maximum of 4 days after randomisation.
Neuropsychiatric assessment	Baseline and post infusion: Coma grade at baseline (Table 3); Modified Glasgow Coma Scale (Table 3) assessed at 10 minutes before and every 10 minutes after the intervention for a maximum of 3 hours; Continuous electroencephalography recorded 15 minutes before and 10 minutes after the infusion (Table 4).
Inclusion period (date)	January 1993 to December 1997.
Country	Italy.
Notes	Included data: serum benzodiazepines were detected in 10 participants (4 with Grade III and 6 with Grade IVa coma). The published paper provides no information about the distribution of these participants to flumazenil or placebo during the first intervention period. Therefore, we were unable to exclude these participants from the analyses. The trial reported on several serious adverse events (Table 7).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated sequential list of block‐randomised assignments.
Allocation concealment (selection bias)	Low risk	Concealed ampoules of flumazenil and placebo.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Blinding of participants and personnel using placebo.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinding of outcome assessment using placebo.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No missing outcome data and all participants included in analyses.
Selective reporting (reporting bias)	Low risk	Trial described clinically relevant outcomes. We had no access to information about outcomes described in the original protocol or information in trial registries.
For‐profit funding	Low risk	No information provided.
Other bias	Low risk	No other biases.
Overall assessment	Low risk	Low risk of bias.