BACKGROUND: Acute non-variceal upper gastrointestinal bleeding (NVUGIB) is a common cause of admissions as well as aggressive transfusion of blood products. Whether the transfusion strategy in NVUGIB impacts on hemostasis is unknown and constitutes the focus of this study.

METHOD: Retrospective analysis of all hospital admissions in Denmark between 2011 and 2013 where hemostatic endoscopic interventions in either the stomach or duodenum had been employed. Regression modeling was used to predict the effect of units transfused of packed red blood cells (PRBC), fresh frozen plasma (FFP), and platelets (PLT) on primary outcome 30-day mortality as well as secondary hemostasis-related outcomes and need for re-endoscopy and conversion to surgery. The model was corrected for confounders, including transfusion of other blood products (PRBC, FFP, and PLT, respectively), patient age as well as pre-existing medical conditions.

RESULTS: 5107 patients received 10783 therapeutic endoscopic interventions. Units of PRBC transfused were identified as a predictor of re-endoscopy, surgery, and 30-day mortality with odds ratio (OR) 1.08 (1.06-1.09, p < 0.01), 1.05 (1.03-1.07, p < 0.01), and 1.04 (1.01-1.06, p < 0.01), respectively. Units of FFP transfused were associated with a higher risk of surgery and 30-day mortality with OR 1.05 (1.02-1.08, p < 0.01) and 1.04 (1.02-1.07, p < 0.01), respectively. Units of PLTs transfused were independently associated with a reduction in risk of re-endoscopy 0.93 (0.87-0.98, p = 0.02). A high ratio of PRBC:FFP:PLT (1:1:1) was associated with reduced need for re-endoscopy OR 0.23 (0.06-0.67, p = 0.01) but increased mortality with OR 3.60 (1.34-11.38, p = 0.02).

CONCLUSION: PRBC transfusion was associated with adverse events, including 30-day mortality and failure of hemostasis. In contrast, transfusion of PLT was associated with a reduction in need for re-endoscopy.