Predictive Factors for Endoscopic Visibility and Strategies for Pre-endoscopic Prokinetics Use in Patients with Upper Gastrointestinal Bleeding

Background Although current guideline recommends selective use of pre-endoscopic prokinetics to increase diagnostic yield in upper gastrointestinal bleeding (UGIB) patients, no data to guide the use of these drugs are available. Aims We aimed to investigate predictive factors for endoscopic visibility and develop simple and useful strategies for pre-endoscopic prokinetics use in UGIB patients. Methods A total of 220 consecutive patients who underwent upper endoscopy for suspicious UGIB were enrolled. Patients were randomly allocated to either a training or a validation set at a 2:1 ratio. Significant parameters on univariate analysis were subsequently tested by a classification and regression tree (CART) analysis. Results Time to endoscopy and nasogastric aspirate findings were independently related to endoscopic visibility. The CART analysis generated algorithms proposed sequential use of time to endoscopy (<= 5.2 vs. >5.2 h) and nasogastric aspirate findings (red blood or coffee rounds vs. clear aspirate) for predicting endoscopic visibility. Prediction of unacceptable visibility in the validation set produced sensitivity, specificity, positive predictive value, and negative predictive value of 75.8, 67.5, 65.8, and 77.1 %, respectively. Accurate prediction for visibility was identified in 52 of 73 patients (71.2 %). Conclusions Time to endoscopy and nasogastric aspirate findings were independently related to endoscopic visibility in patients with UGIB. A decision-tree model incorporating these two variables may be useful for selecting UGIB patients who benefit from pre-endoscopic prokinetics use.