Betting on the score: results of a probe▸

The Rockall score has proved useful in predicting who is at higher risk of recurrent bleeding or death in upper gastrointestinal haemorrhage, and is simple to use. It was developed, however, in unselected patients, most of whom did not undergo endoscopic therapy. The authors prospectively assessed the validity of the Rockall score in patients undergoing heater probe with or without thrombin injection therapy for peptic ulcer bleeding. A cohort of 247 patients who took part in a randomised trial of therapy (previously published in Gastroenterology 2003;125:396–403) formed the study group and Rockall score was prospectively collected. The aim was to assess the utility of the score in high risk patients undergoing endotherapy.

Obviously the score was high (mean 5.8) in all patients; 15% had recurrent bleeding in the trial and mortality was 9%. In terms of calibrating the score, predicted probabilities of recurrent bleeding and mortality were both significantly higher than those actually observed …