Seven trials (n=936) were included in the review, comprising five fully published trials and two abstracts. Sample sizes ranged from 31 to 359 patients. Three trials were assigned Jadad quality scores of 1 point, two trials scored 3, one trial scored 4 and one trial scored the maximum Jadad score of 5 points.
There were no significant differences between proton pump inhibitors and histamine-2 receptor antagonists in stress-related upper gastrointestinal bleeding prophylaxis (seven trials; n=936 patients, I2=66%), pneumonia (six trials; n=905 patients; I2=0%) and mortality (three trials; n=569 patients; I2= 0%) in patients admitted to intensive care units. Removal of one trial reduced the between-trial heterogeneity for the analysis of stress-related upper gastrointestinal bleeding (I2=26%).
There were no differences between proton pump inhibitors and histamine-2 receptor antagonists in subgroups analyses that evaluated the requirement for mechanical ventilation as an inclusion criterion, reporting of baseline characteristics, route of administration, trial quality, whether upper gastrointestinal bleeding was well defined or otherwise, and the type or dose of proton pump inhibitor. Trials published prior to 2000 showed significant benefits with proton pump inhibitor treatment (RD -0.11, 95% CI -0.21 to -0.01, three trials; I2=54%; supplementary details published online, see URL for additional data).
The funnel plot showed some evidence of publication bias, language bias and inflated estimates of effect because of flawed methodology in smaller included trials.