Validation of the Blatchford Bleeding Score in Cancer Patients with Upper Gastrointestinal Bleeding

Shin, Ju Yong; Ahn, Shin; Lee, Youn Sun; Lim, Kyoung Soo

Abstract

PURPOSE
The Glasgow Blatchford Score (GBS) and the Rockall score are validated risk tools for predicting adverse outcomes in patients with upper gastrointestinal bleeding (UGIB). In this study, we attempted to validate these risk assessment tools in patients with active cancer who visited an emergency department (ED) with UGIB.
METHODS
We retrospectively reviewed electronic medical records of patients with active cancer presented to Asan Medical Center ED from January 2009 to December 2011. The primary outcomes required therapeutic interventions (transfusion, endoscopic/surgical/radiologic interventions), and there was a recurrence of bleeding or mortality within 30 days.
RESULTS
Of the 225 patients, 197(87.6%) needed interventions. The area under the receiver-operator curves showed that the GBS [0.86; 95% Confidence Interval (CI), 0.77-0.95] surpassed the clinical Rockall (0.67; 95% CI, 0.55-0.79) and full Rockall scores (0.72; 95% CI, 0.61-0.83) in predicting clinical interventions. Regarding a score of 2 or less as negative, the GBS showed a sensitivity of 0.99 and a specificity of 0.54. When patients were divided according to their source of bleeding, the sensitivity and specificity did not change.
CONCLUSION
The GBS outperformed clinical and full Rockall scores in predicting the intervention in patients with active cancer. The source of bleeding was not an important factor in the score's performance. The GBS also showed very good sensitivity; however, its specificity is suboptimal and limits its role as a sole indicator for decisions in cancer patients with UGIB.