[Can early recurrent hemorrhage in gastroduodenal ulcer be predicted?].

Objective: To evaluate clinical, biological and endoscopic predictive factors of early recurrent bleeding from gastric or duodenal ulcers.

Patients And Methods: Seventy six patients (26 females and 50 males) with a mean age of 65.9 years (SD = 17.2) were included in a prospective study. Among the 76 patients, there were 39 NSAIDs users (51.3%). An endoscopy was performed systematically until 12 hours after admission. Treatment modalities were identical for all patients. The predictive value of 11 factors was evaluated through an univariate and multivariate analysis.

Results: Three factors had independent significant predictive value: a) the number of blood units used to treat a shock and to increase the haemoglobin level up to 100 g/L (P < 0.05); b) "high endoscopic risk" of recurrent bleeding including Forrest Ia, IIa and IIb ulcers (P < 0.05); c) a non steroidal anti-inflammatory treatment was associated with no recurrent bleeding contrary to the other factors (P < 0.05). Based on a second multivariate analysis including clinical factors only, a predictive score was calculated: 5 + number of blood units -5 x (NSAID = 0/1). The cut off point with maximum discrimination was 6 (specificity = 79.6%; sensitivity = 77.2%).

Conclusion: A combination of clinical and endoscopic factors is useful to predict ulcer recurrent bleeding. Our clinical predictive score is interesting because of its simplicity. Its predictive value is of interest but have to be evaluated in another sample of patients.