Table 1.

Strategies to minimise bias

Stage and type of bias	Strategy implemented
Protocol development
Design bias	Extensive international scientific, clinical and ethical input on the protocol; patient and family input to refine the patient-important bleeding outcome
Corporate conflicts of interest	Peer-review funded trial; locally sourced pantoprazole
Procedural bias	Standard operating procedures guide protocol implementation; central statistical monitoring is ongoing throughout the trial
Omission bias	Eligibility criteria are broad; enrolment is in five continents
Surveillance bias	Rigorous training of research personnel
Detection of ventilator-associated pneumonia (VAP)	To avoid biased choice of VAP definition: VAP reporting has one main and several alternate definitions
Protocol implementation
Prognostic imbalance	At point of randomisation, patients are stratified for prehospital acid suppression which may influence outcomes
Selection bias	Allocation is concealed; research personnel screening, consenting, and enrolling patients are unaware of randomisation sequence
Detection and performance bias	Patients, families, all clinical and research personnel are blinded
Measurement bias	Primary efficacy outcome: Clinically important GI bleeding is centrally adjudicated by two physicians trained in study procedures, and blinded to allocation and centre
Loss to follow-up	Primary Safety Outcome: for 90-day mortality status, multiple methods used for patients discharged alive before 90 days; all other outcomes are hospital based as recorded in medical charts
Missing data	Each research record is reviewed and validated at least three times by Methods Center staff
Analysis
False claims of benefit	A priori statistical approach is very conservative for stopping early for apparent benefit before full sample size reached
False claims of no difference	A priori statistical approach does not include stopping early for futility before full sample size reached
Confirmation bias	Analyst is blinded to allocation until after the final analysis
Analytic bias	Analysis will adhere to the intention-to-treat principle
Dissemination
Reporting bias	Trial reporting will adhere to trial registration (NCT03374800), protocol and statistical analysis plan
Publication bias	Results will be disseminated through many knowledge translation strategies including peer-review journals

These are the strategies we protocolised to minimise bias in four different phases of the trial.