Prediction of further hemorrhage in bleeding peptic ulcer

To obtain a simple mathematical model able to estimate early the risk of further hemorrhage in bleeding peptic ulcer.

A prospective study was conducted on 1567 patients admitted for acutely bleeding peptic ulcer. Ten readily available variables were tested for association with further hemorrhage, and then a logistic regression analysis was carried out.

Further hemorrhage occurred in 312 (20%) patients. The univariate analyses showed age over 65 yr, male sex, alcoholism, associated disease, hematemesis, endoscopic bleeding stigmata and shock, and blood urea over 90 mg/dl at admission to be significantly associated with continuous hemorrhage or rebleeding (p < 0.05 to p < 0.001). When these were assessed in a multivariate logistic regression analysis, the best fitted model (likelihood ratio test = 9.9; p = 0.5) included four independent variables. Bleeding stigmata (p < 0.001; odds ratio = 3.1), shock (p < 0.001, odds ratio = 2.5), hematemesis (p < 0.001; odds ratio = 1.6) and age over 65 (p = 0.04); odds ratio = 1.3) were associated with high risk of further hemorrhage.

Bleeding stigmata and changes in hemodynamics are confirmed as the most relevant predictors of further hemorrhage. When they are considered with hematemesis and age altogether, other clinical features loss importance for prognosis. By taking advantage of the joint influences of these variables, accuracy of prognosis improves with respect to that obtained when they are considered as individual risk factors.