van Essen 1985.

Methods	Open‐label randomised controlled trial
Participants	Baseline characteristics Number randomised: 90 participants Number analysed: 58 participants Prostaglandin Age (years; median (SD)): ‐ (‐) Number of participants (n): 29 Gender (n, male/female): ‐ Placebo Age (years; median (SD)): ‐ (‐) Number of participants (n): 29 Gender (n, male/female): ‐ Inclusion criteria Admitted to surgical or medical ICU Exclusion criteria Previous oesophageal or gastric surgery Presence of oesophageal varices or extensive burns Evidence of GI bleeding on ICU admission Expected discharge or expected oral feeding in 3 days Baseline imbalances: Quote: "29 participants in both the groups were similar with regard to sex distribution, age and number and nature of risk factors" Comment: no significant difference between the 2 groups with respect to demographic and baseline risk factors
Interventions	Prostaglandin Dose (total/d): ‐ Duration of treatment (days, mean (SD)): 3 ‐ 7 (‐) Route: gastric tube Intervention: intragastric prostaglandin E2PGE2; 0.5 mg dissolved in 20 mL of water + flushed with 20 mL water, administered every 4 hours via gastric tube Concomitant medications: Placebo Dose (total/d): ‐ Duration of treatment (days, mean (SD)): 3 ‐ 7 (‐) Route: gastric tube Intervention: only water, flushed with 20 mL of water, administered every 4 hours via gastric tube Concomitant medications: ‐ Adherence to regimen: Quote: "The study lasted 3 ‐ 7 days. Treatment was discontinued if enteral feeding was given within the first three days, if gastric tube was removed or participant underwent surgery or had GI bleeding" Of 32 participants, 8 participants had gastrectomy performed within first 3 days; in 1 participant, no chromium labelling was performed, 6 participants were discharged within first 3 days from ICU 3 participants showed non‐compliance with entry criteria, 12 participants received enteral feeds or underwent gastric tube removal within first 3 days, and 9 participants died within the first 3 days Duration of trial: November 1981 to September 1983 Duration of follow‐up: ‐
Outcomes	Outcomes sought in review and reported in trial Primary outcomes GI haemorrhage measured every 3.5 hours by inspection for manifest haemorrhage and by peroxidise test based on orthotolidine reaction and qualitatively measured every 24 hours by Cr–chromate labelling for erythrocytes. More than 15 mL blood loss from the upper GI tract was considered to be evidence of mucosal damage in the stomach Secondary outcomes All‐cause mortality in ICU Outcomes sought but not reported in trial Participants requiring blood transfusion Units of blood transfused Nosocomial pneumonia All‐cause mortality in hospital Length of stay in ICU Length of stay on ventilator Adverse events Outcomes reported but not used in review Nil
Notes	Setting: Departments of Internal Medicine and Surgery, University Hospital, Dijkzigt, Rotterdam, The Netherlands Source of funding: Quote: "The study was supported, in part, by a grant–in‐aid from Upjohn Inc., Kalamazoo, MI" Conflicts of interest: ‐ Ethics approval: ‐ Informed consent: ‐ Clinical trials registration: ‐ Sample size calculation: ‐
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: not clearly mentioned in the study report
Allocation concealment (selection bias)	Unclear risk	Comment: not clearly mentioned in the study report
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Patients were randomly assigned to receive placebo or prostaglandin therapy in a double blind fashion" Comment: This was a placebo‐controlled trial where most likely participants and study personnel were blinded to the interventions. Therefore, the likelihood of performance bias is low
Blinding (detection bias) Clinically important upper GI bleeding	Low risk	Comment: unclear on blinding of outcome assessors. However, GI bleeding was an objective outcome that was detected as per the definition in the study protocol. Therefore the likelihood of detection bias is low
Blinding (detection bias) Nosocomial pneumonia	Unclear risk	Comment: Study did not address this outcome
Blinding of outcome assessment (detection bias) Adverse reactions of interventions	Low risk	Comment: unclear on blinding of outcome assessors. However, the outcome of interest was objective in nature, so the likelihood of detection bias is low
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: Not all randomised participants were part of the final analysis. 90 were randomised, and only 58 were part of the study. However, this was a per‐protocol analysis, and the numbers were well balanced between groups. Therefore there is no serious attrition bias
Selective reporting (reporting bias)	Low risk	Comment: All intended outcomes were analysed and reported in the study
Other bias	Low risk	Comment: Upjohn Inc., Kalamazoo, MI, partially funded the study. The role of the sponsor in the conduct and reporting of the trial is unclear. No other sources of bias are suspected