Abstract
BACKGROUND: Several prognostic scales have been developed for use in upper gastrointestinal bleeding (UGIB), including the Glasgow-Blatchford score, which uses blood urea nitrogen (BUN) as one of eight prognostic variables. However, the test characteristics of BUN in the identification of UGIB or high-risk endoscopic lesions have not been clearly determined. AIMS: This study aimed to evaluate if BUN independently predicts the presence of positive endoscopic findings in cases of presumed UGIB and determine a threshold urea level above which it is more likely to identify a source of UGIB on endoscopy. METHODS: A crude odds ratio was calculated for odds of bleeding being identified on upper endoscopy based on thresholds of urea>=5, 7.5, 10, 12.5 and 15, compared to values lower than these. Adjusted odds ratios were then calculated using logistic regression to account for the factors that were determined a priori. Covariates included in the model were age, sex, hemoglobin, presence of melena, presence of hematemesis, admission in ICU, and use of ASA, warfarin, clopidogrel, or NSAIDS. RESULTS: The odds of identifying UGIB at endoscopy for patients with a urea >=10 was 3.73 (95% CI: 1.90–7.31) times higher than for patients with urea < 10. Variables that were significantly associated with identifying source of bleeding at upper GI endoscopy included male gender and symptoms of melena or hematemesis, after adjusting for the impact of other covariates. BUN >20 is predictive of UGIB in the following settings: normal renal function (spec 98%, PPV 0.86), melena (spec 81%, PPV 0.8), NSAID+anticoagulation/ASA use (spec 85%, PPV 0.8), and cirrhotic patients (spec 100%, PPV 1.0). BUN >20 is also predictive of positive EGD findings (spec 87%, PPV 0.8) and high risk lesions (spec 81%, NPP 0.83). A BUN level >20 had 82% specificity for high risk endoscopic lesion requiring intervention, but lower BUN levels were not able to predict EGD intervention. BUN >15 is predictive of UGIB in hematemesis (spec 95%, PPV 0.95) or NSAID users (spec 100%, PPV 1.0). BUN levels >20 had a specificity of 87% for UGIB but poor sensitivity (23%), in contrast to the Glasgow-Blatchford score which is highly sensitive (>90%) but poorly specific (<20%). CONCLUSIONS: Overall, the results of this study provide new clinically relevant information regarding the operating characteristics of BUN for UGIB. In males, patients with normal renal function, cirrhosis, NSAID/ASA/AC users, or symptoms of melena and hematemesis, a high BUN level is predictive of positive endoscopic findings in presumed UGIB. A BUN level >20 predicts the need for endoscopic intervention but levels below 20 do not correlate well with the need for endoscopic intervention. BUN level alone is more specific for UGIB when compared to the Glasgow-Blatchford score, which has a higher sensitivity. FUNDING AGENCIES: None