Background: Haematemesis is an alarming symptom in children, even if the proportion of normal endoscopies ranges from 10 to 21% and the causes are often benign. The purpose of the study was to identify clinical predictors of endoscopic lesions with high risk of bleeding and to establish a score that predict the presence of these lesions.
Methods: Retrospective study carried in Children's Hospital of Tunis between 1997 and 2006 involved children with haematemesis who underwent Upper gastrointestinal endoscopy. Several clinical parameters were analyzed. Univariate analysis and multivariate logistic regression were performed to identify predictive parameters of endoscopic lesions with high risk of bleeding. A score was developed from the parameters derived from the multivariate analysis. The sensitivity and specificity of the score were determined.
Results: Among 2814 endoscopies, 814 were conducted for haematemesis and 489 were selected for the study. 140/489 had lesions with high risk of bleeding. Multivariate logistic regression analysis identified six factors independently associated with high risk bleeding lesions: endoscopy performed within 48 hours (OR=2.2; 95% CI 0.7-6.9), re-bleeding (OR=1.4; 95% CI 0.7-2.5), the importance of the bleeding, mild to severe (OR=1.8; 95% CI 1.1-3), bright red haematemesis (OR=1; 95% CI 0.2-5.8), history of gastrointestinal and liver disease (OR=1.6; 95% CI 1.1-3) and intake of gastro toxic drugs (OR=1.3; 95% CI 0.8-2.3). Then, we established a score. The sensitivity, specificity, positive predictive value and negative predictive value of this score were respectively 79.6%, 32.9%, 34.9% and 78% for a cut off value>0.22.
Conclusion: The clinical predictive parameters of high risk bleeding lesions identified have not yielded a score with significant sensitivity and specificity. A prospective study should be performed to improve the score.