Model to predict rebleeding following endoscopic therapy for non-variceal upper gastrointestinal hemorrhage

Following endoscopic therapy, up to 20% of patients with non-variceal upper gastrointestinal hemorrhage experience rebleeding. The aim of the present study was to determine risk factors for recurrent hemorrhage in these patients.

This was a retrospective cohort study of consecutive patients admitted to a tertiary care hospital between 1 July 1999 and 30 June 2004, with non-variceal upper gastrointestinal hemorrhage. Patients were evaluated for rebleeding within 30 days of successful therapeutic endoscopy. Using the hospital's endoscopic database, 236 patients were identified. Risk factors were identified using multivariable logistic regression with backward selection. Internal validation was carried out using bootstrapping.

Six risk factors were identified: failure to use a proton pump inhibitor post-procedure (P = 0.056), Endoscopically demonstrated bleeding (P = 0.053), peptic ulcer as the bleeding source (P = 0.018), treatment with epinephrine monotherapy (P = 0.0026), post-procedure intravenous or low molecular weight heparin use (P = 0.0014), and moderate or severe cirrhosis (P = 0.032) (PEPTIC). The risk of rebleeding increased as the number of risk factors present increased. The observed rates of rebleeding were: 7.1%, 16.4%, 37.0%, 75.0% and 100% for zero, one, two, three or four risk factors, respectively (no patients had five or six risk factors present). The bias-adjusted area under the receiver-operator characteristic curve for the number of risk factors predicting rebleeding was 0.69.

We have identified six easily remembered risk factors, which, when summed, predict recurrent hemorrhage following endoscopic therapy for non-variceal upper gastrointestinal hemorrhage.