Background/aims: : Risk scoring systems for upper gastrointestinal (UGI) bleeding have not been well validated for tumor bleeding. This study aimed to identify risk factors for mortality in patients with UGI cancer bleeding and to develop a predictive model.
Methods: : Consecutive patients with UGI cancers who underwent esophagogastroduodenoscopy for suspected bleeding were retrospectively included. Patient characteristics, endoscopic findings and 30-day mortality were assessed. A predictive model was made based on risk factors for mortality using logistic regression, and the area under the curve (AUC) of this model was calculated. It was then compared with other risk scoring systems.
Results: : In a total of 264 patients, 193 had tumor bleeding. Among them, 108 (56.0%), 76 (39.4%), and nine (4.7%) patients received conservative treatment, endoscopic therapy, and non-endoscopic hemostasis, respectively. Rebleeding occurred in 23 (21.3%), 26 (34.2%), and one (11.1%) patient(s), respectively. Our new model is composed of altered mental status, renal failure, rebleeding, age older than 65 years, and low serum albumin (all p<0.05). This model predicted 30-day mortality with an AUC of 0.79 (95% confidence interval, 0.72 to 0.86), which was significantly higher than AUCs of the Glasgow-Blatchford score, Rockall, and AIMS65 score (AUC=0.61, 0.64, and 0.69, respectively, all p<0.05).
Conclusions: : Our new scoring system provides a better prediction of 30-day mortality than existing scoring systems in patients with UGI cancer bleeding. This new scoring system can be used to predict and prepare these patients who are known to have high mortality.
Keywords: Gastrointestinal hemorrhage; Gastrointestinal neoplasm; Mortality; Prognosis; Risk assessment.