Epidemiology of gastrointestinal bleeding in adult patients on extracorporeal life support.

Abstract

Dear Editor, Gastrointestinal (GI) bleeding is associated with high mortality in critically ill patients [1]. Stress ulcer prophylaxis has reduced its incidence; however, some patients remain at risk. There are few data about GI bleeding during adult extracorporeal life support (ECLS).We performed a retrospective cohort study to determine its incidence and impact on mortality. The institutional review board approved our study. A total of 132 patients were included over 3.5 years. Fifty-four patients had veno-arterial (VA) ECLS, 64 patients had veno-venous (VV) ECLS, and 14 patients had sequential therapy. The median number of ECLS days was 7. GI bleeding was defined as hematemesis, melena, or hematochezia. Anticoagulation was with heparin and anticoagulation goals were an activated clotting time (ACT) of 180–200 s or activated partial thromboplastin time (aPTT) of 60–80 s for VA ECLS and an ACT of 140–160 s or aPTT of 45–55 s for VV ECLS. Eighteen patients (13.6 %) experienced GI bleeding. These patients had longer ECLS courses (12 days vs. 7 days, p = 0.002) and higher inhospital mortality (83.3 vs. 43.9 %, p = 0.002) (Table 2 in the supplementary material). The median presentation day was 8 with most patients (n = 10, 55.6 %) having coffee ground emesis. Details of the bleeds are listed in Table 3 in the supplementary material. Esophagogastroduodenoscopywas performed in seven patients and the most frequent diagnosis was stress gastritis. Most patients were managed with transfusion, hematocrit monitoring, and proton pump inhibitors. Three patients had their anticoagulation temporarily held. GI bleeding was independently associated with in-hospital mortality (OR 5.91; 95 % CI 1.44, 24.25; p = 0.01) after controlling for important confounders (Table 1). GI bleeding may be more frequent in ECLS patients as a result of multifactorial coagulopathy, generation of systemic inflammation, and nonpulsatile blood flow during VA ECLS, which can lead to reduced gastric perfusion and pH [2–5]. Future prospective studies would be helpful in confirming our findings.