PURPOSE: To systematically identify predictors of gastrointestinal (GI) bleeding in adult intensive care unit (ICU) patients.

METHODS: We conducted a systematic review and meta-analysis of cohort studies including trial cohorts. We searched MEDLINE, EMBASE, and trial registries up to March 2019. Eligible studies assessed potential predictors of clinically important GI bleeding (CIB; primary outcome) or overt GI bleeding (secondary outcome), had > 20 events, and presented adjusted effect estimates. Two reviewers assessed study eligibility, extracted data, and assessed risk of bias and certainty of evidence using GRADE. We meta-analysed adjusted effect estimates if data from ≥ 2 studies were available.

RESULTS: We included 8 studies (116,497 patients). 4 studies (including 74,456 patients) assessed potential predictors of CIB, and we meta-analysed 12 potential predictors from these. Acute kidney injury (relative effect [RE] 2.38, 95% confidence interval [CI] 1.07-5.28, moderate certainty) and male gender (RE 1.24, 95% CI 1.03-1.50, low certainty) were associated with increased incidence of CIB. After excluding high risk of bias studies, coagulopathy (RE 4.76, 95% CI 2.62-8.63, moderate certainty), shock (RE 2.60, 95% CI 1.25-5.42, low certainty), and chronic liver disease (RE 7.64, 95% CI 3.32-17.58, moderate certainty) were associated with increased incidence of CIB. The effect of mechanical ventilation on CIB was unclear (RE 1.93, 0.57-6.50, very low certainty).

CONCLUSIONS: We identified predictors of CIB and overt GI bleeding in adult ICU patients. These findings may be used to identify ICU patients at higher risk of GI bleeding who are most likely to benefit from stress ulcer prophylaxis.