World Congress at ACG2017
Simultaneous Plenary Session 4A: Liver
61 - Early Prediction of Variceal Hemorrhage in Upper Gastrointestinal Bleeding
Wednesday, October 18
9:50 AM - 10:00 AM
Location: Valencia Ballroom BC (Level 4)
Brandon Teng, MD
University of Washington Medical Center
Category: GI Bleeding
Brandon J. Teng, MD1, Stephen Bach2, Anand C. Baxi, MD1, Michael A. Harms, MD1, Elsbeth Jensen-Otsu, MD1, Cynthia W. Ko, MD, MS2, Lisa L. Strate, MD, MPH, FACG3
1University of Washington Medical Center, Seattle, WA; 2University of Washington, Seattle, WA; 3University of Washington School of Medicine, Seattle, WA
Introduction: Variceal hemorrhage remains a major cause of morbidity and mortality for patients with upper gastrointestinal bleeding (UGIB). Our aim is to identify clinical and laboratory variables associated with variceal hemorrhage in order to develop a clinical prediction model of variceal versus non-variceal bleeding.
Methods: We performed a retrospective chart review, identifying patients admitted with UGIB at a university and county hospital between 2010 and 2015. Patient outcome was determined by variceal or non-variceal bleeding source on endoscopy. Patients were excluded if they were transferred from another hospital, did not receive an esophagogastroduodenoscopy (EGD), or did not have an UGIB source identified. Statistical analysis was performed using least absolute shrinkage and selection operator (LASSO) logistic regression to identify variables present on admission associated with variceal bleeding. Ten-fold cross validation with area under the curve was used to identify the number of variables to include and to calculate the magnitude of their B estimates with the ideal penalty tuning parameter. We developed a prediction score based on these factors and compared this with the physician impression of possible variceal bleeding.
Results: A total of 449 patients were included in the study, with 84 (18.7%) presenting with UGIB from a variceal source. 114 patients had known cirrhosis prior to admission, of which 66 (57.9%) had a variceal source of bleeding. Five factors were associated with variceal bleeding, including platelets less than 150 thousand/µL (exp(ß)=3.11), bilirubin greater than 2 mg/dL (exp(ß)=1.56), INR greater than 1.5 (exp(ß)=1.41), AST greater than 34 units/L (exp(ß)=1.34) and ALT greater than 40 units/L (exp(ß)=1.34). A predictive score for etiology of UGIB was developed using these factors. A score of greater than -2.2 was associated with a variceal source, and correlated with a sensitivity of 96%, specificity of 38%, negative predictive value of 98%, and positive predictive value of 27%. Subgroup analysis of patients with cirrhosis did not reveal any variables associated with a variceal source of bleeding. The prediction model was more sensitive than physician impression of possible variceal bleeding (96% vs. 84%, p=0.007).
Discussion: We identified multiple factors associated with variceal bleeding. Our prediction score may be useful to rule out variceal hemorrhage and inform early management of patients presenting with UGIB.
Supported by Industry Grant: No
Citation: . EARLY PREDICTION OF VARICEAL HEMORRHAGE IN UPPER GASTROINTESTINAL BLEEDING. Program No. 61. World Congress of Gastroenterology at ACG2017 Meeting Abstracts. Orlando, FL: American College of Gastroenterology.