100,000 adults, hospital admissions approaching 400,000 annually, recurrence rates estimated at 20% and mortality approaching 10-14%. Although the early use of a risk stratification score in upper GI bleeding is recommended, existing risk scores are not widely used in clinical practice. The purpose of this study was to construct a simple bedside prediction tool that accurately identifies patients at increased risk for in-hospital mortality after NVUGIB without the need for endoscopic parameters. Methods: This was a retrospective study at a tertiary care referral center between January 2008 and November 2011. All patients O 18 years of age admitted with a diagnosis of acute NVUGIB were included. Rebleeding was defined as NVUGIB following index endoscopic therapy within 30 days of an initial bleed. Demographics, presentation, comorbidities, medications, serum and hemodynamic parameters were collected. Multivariable logistic regression analysis was performed to build a model for prediction of rebleeding. Accuracy to predict mortality was assessed by the area under the receiver operating characteristic (AUROC) curve. Results: A total of 251 patients underwent EGD for NVUGIB (mean age 64 13 years, 39% female, 74% Caucasian). The most common source of bleeding was peptic ulcer disease (42%). A total of 123 (49%) patients had a recurrent bleed after index endoscopy. The average time to an episode of rebleeding was 3.5 6.1 days. The majority of patients only had 1 episode of rebleeding (58%). Average length of ICU stay was 15 days. The in-hospital mortality rate was 18% for all NVUGIB. The 4 factors present on admission with the best discrimination for mortality risk were age, platelets, WBC and diastolic blood pressure as depicted in the nomogram in Figure 1. The model had a high predictive accuracy (AUCROC Z 0.76; Figure 2). Conclusions: Efforts to identify patients at highest risk for recurrent non-variceal upper GI bleeding mortality are of significant clinical utility. This simple, accurate risk score reliably predicts with 76% accuracy in-hospital mortality in patients with recurrent non-variceal upper gastrointestinal bleeding. Figure 1. Nomogram for prediction of ICU mortality. Figure 2. Receiver Operating Characteristics (ROC) plot for prediction of ICU mortality. AB298 GASTROINTESTINAL ENDOSCOPY Volume 79, No. 5S : 2014 Sa1677 Risk Scoring Systems in Predicting Intervention and Clinical Outcomes of Bleeding in Patients With Unresectable Gastric Cancer Young-IL. Kim*, IL Ju Choi, Soo-Jeong Cho, Jong Yeul Lee, Chan Gyoo Kim, MI–Jung Kim, Hark Kyun Kim, Young Iee Park Center for Gastric Cancer, National Cancer Center, Korea, Gyeonggi-do, Republic of Korea Background & Aims: Pre-endoscopic Rockall score (RS), full RS, and Glasgow-Blatchford bleeding Score (GBS) are increasingly used to stratify the risk in patients with upper gastrointestinal bleeding (UGIB). The aim of this study was to evaluate the efficacy of those bleeding risk scoring systems in patients with UGIB due to unresectable advanced gastric cancer (AGC). Methods: We reviewed retrospectively the medical records of patients presenting with UGIB due to unresectable AGC at the National Cancer Center, Korea from May 2001 to June 2012. Pre-endoscopic RS, full RS and GBS were calculated. The area under the receiver operating characteristicscurve (AUC) was used to assess the performance of these scores to predict the need for interventions and clinical outcomes. Results: During the study period, 355 patients who presented with UGIB from unresectable AGC received endoscopy. Of these, interventions were needed in 118 (33.2%). A total of 115 patients (32.4%) underwent endoscopic therapy, 2 (0.6%) required transarterial embolization and one (0.3%) underwent surgery as an initial treatment. Full RS was useful to predict the need for intervention (AUC 0.77, P! 0.001) and 7-day mortality after bleeding (AUC 0.67, PZ 0.024). However, pre-endoscopic RS, full RS and GBS were not useful for the prediction of rebleeding after initial hemostasis (AUC 0.49, 0.54 and 0.54, respectively), the need for admission more than 3 days (AUC 0.57, 0.64 and 0.62, respectively) and 30-day mortality (AUC 0.51, 0.53 and 0.59, respectively). Conclusions: Full RS might be useful in predicting the need for interventions and 7day mortality of patients presented with UGIB due to unresectable AGC. Sa1678 Prevalence of Heyde’s Syndrome: Does It Just Pertain to Aortic Stenosis? a Case-Control Study Karen Draper, Robert J. Huang, Lauren B. Gerson* Gastroenterology, California Pacific Medical Center, San Francisco, CA; Medicine, Stanford University, Stanford, CA Background: Heyde’s syndrome is described as the relationship between aortic stenosis (AS) and gastrointestinal angiodysplastic lesions (GIAD). While prior studies have examined the prevalence of AS in patients with GIAD-associated GIB, there is little data regarding whether this syndrome exists in patients with other cardiac valvular disorders. Our study aimed to determine the prevalence of Heyde’s syndrome in patients with AS compared to a control group of patients with mitral regurgitation by examining rates of iron deficiency anemia and GIAD-associated GIB in these two groups. Methods: We used an echocardiography database at a single large academic medical center to identify adult patients with severe AS and moderate or severe mitral regurgitation (MR). Patients were excluded if they had active malignancy, hematologic conditions, end-stage renal disease, inflammatory bowel disease, history of colectomy, cirrhosis, or if they had no available lab values within 2 years of the echocardiogram. Records from identified patients were then reviewed to assess for the presence of anemia, iron deficiency, and/or overt gastrointestinal bleeding (GIB). GIB was defined as melena, hematochezia, hematemesis, or guaiac positive stool. Values for hemoglobin (Hb), mean corpuscular volume (MCV), and iron panel results from within 2 years of echocardiography were collected. We searched the endoscopy database to determine which patients underwent subsequent endoscopic procedures and location of findings. Estimating that 10% of patients with AS would have GIAD compared to 3% of patients with MR (similar to population controls), the estimate sample size was 152 in each group, using an alpha of 95% and power of 80%. Results: Out of 514 patients who were screened, a total of 298 patients were included in the study (142 MR, 156 AS). Patients with MR were younger and more likely to have congestive heart failure, but otherwise had similar co-morbidities (Table 1). The rates of anemia and microcytic anemia were similar in both groups. The MR group was more likely to have been evaluated for iron deficiency, however rates of confirmed iron deficiency were similar in the two groups. Overall incidence of GIB was similar between the two groups (MR 7.7% vs. AS 8.3%); however, rates of proven GIAD (MR 0% vs. AS 1.3%) and occult GIB (MR 36% vs. AS 54%) appeared higher among patients with aortic stenosis, though again not reaching statistical significance (Table 2). Conclusions: While the prevalence of AS in patients with obscure or GIAD-associated GIB has been shown in previous studies to be relatively common, our study did not suggest that the incidence in AS was greater compared to a control group with MR. Whether MR itself could predispose to GIADassociated GIB is a topic for further research. Table 1 Characteristics of the Study Patients Characteristics Severe AS (N[156) Moderate-Severe MR (N[142) www.giejourn p value Male 93 (60%) 77 (54%) NS Age 79 13.5 66 17 !0.0001