Predicting mortality in non-variceal upper gastrointestinal bleeders: validation of the Italian PNED Score and Prospective Comparison with the Rockall Score

We sought (i) to validate a new prediction rule of mortality (Progetto Nazionale Emorragia Digestiva (PNED) score) on an independent population with non-variceal upper gastrointestinal bleeding (UGIB) and (ii) to compare the accuracy of the Italian PNED score vs. the Rockall score in predicting the risk of death.We conducted prospective validation of analysis of consecutive patients with UGIB at 21 hospitals from 2007 to 2008. Outcome measure was 30-day mortality. All the variables used to calculate the Rockall score as well as those identified in the Italian predictive model were considered. Calibration of the model was tested using the chi2 goodness-of-fit and performance characteristics with receiver operating characteristic (ROC) analysis. The area under the ROC curve (AUC) was used to quantify the diagnostic accuracy of the two predictive models.Over a 16-month period, data on 1,360 patients were entered in a national database and analyzed. Peptic ulcer bleeding was recorded in 60.7% of cases. One or more comorbidities were present in 66% of patients. Endoscopic treatment was delivered in all high-risk patients followed by high-dose intravenous proton pump inhibitor in 95% of them. Sixty-six patients died (mortality 4.85%; 3.54-5.75). The PNED score showed a high discriminant capability and was significantly superior to the Rockall score in predicting the risk of death (AUC 0.81 (0.72-0.90) vs. 0.66 (0.60-0.72), P0.000). Positive likelihood ratio for mortality in patients with a PNED risk score8 was 16.05.The Italian 10-point score for the prediction of death was successfully validated in this independent population of patients with non-variceal gastrointestinal bleeding. The PNED score is accurate and superior to the Rockall score. Further external validation at the international level is needed.