BACKGROUND/AIMS: The aim of this study was to identify the ability of Glasgow-Blatchford score (GBS) and pre-endoscopic Rockall score (pre-E RS) to predict the occurrence of hypotension in patients with non-variceal upper gastrointestinal bleeding who are initially normotensive at emergency department. METHODS: Retrospective observational study was conducted at Asan Medical Center emergency department (ED) in patients who presented with non-variceal upper gastrointestinal bleeding from January 1, 2011 to December 31, 2013. Study population was divided according to the development of hypotension, and demographics, comorbidities, and laboratory findings were compared. GBS and pre-E RS were estimated to predict the occurrence of hypotension. RESULTS: A total of 747 patients with non-variceal upper gastrointestinal bleeding were included during the study period, and 120 (16.1%) patients developed hypotesion within 24 hours after ED admission. The median values GBS and pre-E RS were statistically different according to the occurrence of hypotension (8.0 vs. 10.0, 2.0 vs. 3.0, respectively; p<0.001). In the receiver operating characteristic curve analysis of hypotension development, the area under the curve of GBS and pre-E RS were 66% and 64%, respectively. The sensitivity and the specificity of GBS using optimal cut-off value were 81% and 46%, respectively, while those based on the pre-E RS were 74% and 46%, respectively. CONCLUSIONS: GBS and pre-E RS were both not sufficient for predicting the occurrence of hypotension in non-variceal upper gastrointestinal bleeding. Development of other scoring systems are needed.