Clinical and endoscopic signs for the prediction of recurrent bleeding from gastroduodenal ulcers.

Objective: To evaluate the prognostic importance of various clinical and endoscopic signs in patients who rebled after endoscopic treatment of gastroduodenal ulcers, and to create a predictive index using regression analysis.

Methods: Retrospective study. Methods: Teaching hospital, Lithuania. Methods: 659 patients whose bleeding peptic ulcers were treated endoscopically (out of a total of 949 with bleeding peptic ulcers seen from 1982-1991). Methods: Endoscopic haemostasis (laser or electrocoagulation, or injection sclerotherapy). Methods: Mortality, incidence of rebleeding, and identification of prognostic factors.

Results: Of the 75 patients who rebled 25 died (33%), compared with 23 of the 584 who did not (4%). 57 patients (76%) rebled within two days of endoscopy, and 70 (94%) within a week. 59 variables from the group of 75 who rebled and from a control group of 100 who did not were analysed by logistic regression to construct a predictive index. The following 12 variables were found to be significantly associated with recurrent haemorrhage: frequency and type of bleeding; length of time between episodes of bleeding at home; shock index on admission; speed of reduction in haemoglobin concentration; size and type of ulcer, and whether it was complicated; endoscopic appearance of ulcer; number of ulcers, sex of the patient, and presence of associated diseases (liver disease, coagulopathy, hypertension, or heart disease). The accuracy of the index varied from 78.4% in the prediction of a high risk of rebleeding to 100% in the prediction of a low risk.

Conclusions: By using the predictive index we have been able to improve our management of patients with bleeding peptic ulcers.